Message processing and resource management method, device and equipment

Abstract

The invention provides a message processing and resource management method, device and equipment. The message processing method comprises the following steps: determining identity information of member equipment in a cloud cluster switch; if the identity information indicates that the member device is a first NCP and receives a first protocol message sent by the server through the first port, adding local port information of the first port to the first protocol message to obtain a second protocol message, and sending the second protocol message to the NCC; if the identity information indicates that the member device is the NCC and receives the second protocol message, determining global port information of the first port based on local port information of the first port; establishing a first forwarding table entry, wherein the first forwarding table entry comprises address information of the server and global port information of the first port; and sending the first forwarding entry to the first NCP. Through the scheme of the invention, the forwarding delay is reduced, the traffic packet loss is reduced, and the service quality is improved.

Description

Message processing and resource management method, device and equipment TECHNICAL FIELD

[0001] The present application relates to the technical field of communication, in particular to a message processing and resource management method, device and equipment. BACKGROUND

[0002] A DC (Data Center) is a facility for centralized storage, processing and distribution of data. The DC includes servers, storage devices, network devices, etc., and is used to provide various computing and storage services. A DCI (Data Center Interconnect) is a facility that connects multiple DCs. The DCI is used to implement high-speed, reliable and secure communication between DCs to meet the high-bandwidth and low-latency requirements of data and applications.

[0003] With the rapid development of the Internet, cloud services, big data and AI technology, and the rapid growth of DC network traffic, this puts a huge pressure on the network devices of the DC network. In particular, the development and application of AIGC (Artificial Intelligence Generated Content) has the characteristics of high computing power. From the generation of text and pictures to the generation of audio and video, the computing power and traffic grow exponentially.

[0004] For network devices of a DC network, the network device includes a main control board, an interface board and a network board. Due to the hardware structure limitation of the network device, i.e., the number of main control boards, interface boards and network boards carried is limited, and the main control board, interface board and network board cannot be expanded. Based on this, when the traffic grows a lot, the network device cannot meet the forwarding requirements of a large amount of traffic, resulting in increased delay, traffic packet loss and other problems.

[0005] SUMMARY

[0006] The present application provides a message processing method, which is applied to a member device in a cloud cluster switch, the member device is an NCC, an NCF or an NCP, and the method comprises:

[0007] determining identity information of the member device in the cloud cluster switch;

[0008] if the identity information indicates that the member device is a first NCP and a first protocol message sent by a server is received through a first port, adding local port information of the first port to the first protocol message to obtain a second protocol message, and sending the second protocol message to an NCC;

[0009] If the identity information indicates that the member device is an NCC and the second protocol packet is received, global port information of the first port is determined based on local port information of the first port; a first forwarding table item is established, the first forwarding table item including address information of the server in the first protocol packet and the global port information of the first port; and the first forwarding table item is sent to the first NCP, so that after the first NCP receives a data packet with a destination address being the address information of the server, the first NCP queries the first forwarding table item based on the destination address to obtain the global port information of the first port, and sends the data packet through the first port.

[0010] The application provides a resource management method, which comprises:

[0011] Target resource data of a cloud cluster switch is acquired, the target resource data including one or more of network element level resource data, network level resource data, application level resource data and service resource data;

[0012] Initial demand resource information of a user is acquired, the initial demand resource information indicating expected occupied resources;

[0013] Target demand resource information of the user is determined based on the target resource data and the initial demand resource information, the target demand resource information indicating actually allocated resources for the user;

[0014] The target demand resource information is sent to the cloud cluster switch, so that the cloud cluster switch allocates available resources for the user according to the target demand resource information when a data packet for the user is received, and processes the data packet based on the available resources.

[0015] The application provides a packet processing device, which is applied to a member device in a cloud cluster switch, the member device being an NCC, an NCF or an NCP, and the device comprising:

[0016] A determining module, a receiving module, a processing module and a sending module;

[0017] The determining module is configured to determine identity information of the member device in the cloud cluster switch;

[0018] If the identity information indicates that the member device is a first NCP;

[0019] The receiving module is configured to receive a first protocol packet sent by a server through a first port;

[0020] The processing module is configured to add local port information of the first port to the first protocol packet to obtain a second protocol packet;

[0021] The sending module is configured to send the second protocol packet to the NCC.

[0022] If the identity information indicates that the member device is the NCC, the sending module is configured to send the second protocol packet to the NCC.

[0023] The receiving module is configured to receive the second protocol packet.

[0024] The processing module is configured to determine global port information of the first port based on local port information of the first port, and establish a first forwarding table item, wherein the first forwarding table item includes address information of the server in the first protocol packet and the global port information of the first port.

[0025] The sending module is configured to send the first forwarding table item to the first NCP, so that, after the first NCP receives a data packet with the address information of the server as a destination address, the first NCP queries the first forwarding table item based on the destination address to obtain the global port information of the first port, and sends the data packet through the first port.

[0026] The present application provides a resource management device, the device comprises:

[0027] The acquisition module is configured to acquire target resource data of a cloud cluster switch, wherein the target resource data includes one or more of network element level resource data, network level resource data, application level resource data and service resource data, and acquire initial demand resource information of a user, wherein the initial demand resource information indicates expected occupied resources.

[0028] The determination module is configured to determine target demand resource information of the user based on the target resource data and the initial demand resource information, wherein the target demand resource information indicates actually allocated resources for the user.

[0029] The processing module is configured to send the target demand resource information to the cloud cluster switch, so that, after the cloud cluster switch receives a data packet for the user, the cloud cluster switch allocates available resources for the user according to the target demand resource information, and processes the data packet based on the available resources.

[0030] The present application provides a member device in a cloud cluster switch, wherein the member device is an NCC, an NCF or an NCP, and the member device comprises a processor and a machine readable storage medium, wherein the machine readable storage medium stores machine executable instructions which can be executed by the processor, and the processor is configured to execute the machine executable instructions to implement the packet processing method in the above examples of the present application.

[0031] The application provides a management device, comprising a processor and a machine readable storage medium, the machine readable storage medium stores machine executable instructions capable of being executed by the processor; the processor is used for executing the machine executable instructions to realize the resource management method of the examples of the application.

[0032] The application provides a computer program product, the computer program product comprises a computer program, the computer program is executed by a processor to realize the packet processing method or the resource management method.

[0033] The application provides a machine readable storage medium, the machine readable storage medium stores machine executable instructions capable of being executed by a processor; wherein the processor is used for executing the machine executable instructions to realize the packet processing method or the resource management method.

[0034] From the above technical solutions, in the embodiment of the application, the cloud cluster switch is deployed in the DC network, and the cloud cluster switch comprises an NCC (Network Cloud Controller, cloud cluster controller), an NCF (Network Cloud Forwarder, cloud cluster forwarding device) and an NCP (Network Cloud Packet, cloud cluster access device). The NCC, the NCF and the NCP are independent devices, and the NCC, the NCF and the NCP in the cloud cluster switch can be expanded. When the traffic increases greatly, the cloud cluster switch can also meet the forwarding demand of a large amount of traffic through expansion, reduce the forwarding delay, reduce the traffic packet loss, and improve the service quality of the DC network. BRIEF DESCRIPTION OF DRAWINGS

[0035] Fig. 1 is a structural schematic diagram of a DC network in an embodiment;

[0036] Fig. 2 is a topological structure schematic diagram of a cloud cluster switch in an embodiment;

[0037] Fig. 3A is a schematic diagram of a cloud cluster platform component in an embodiment;

[0038] Fig. 3B is a schematic diagram of a cloud cluster platform component in an embodiment;

[0039] Fig. 4 is a registration flow schematic diagram of a member device in an embodiment;

[0040] Fig. 5 is a flow schematic diagram of a packet processing method in an embodiment;

[0041] Fig. 6 is a flow schematic diagram of a resource management method in an embodiment;

[0042] Fig. 7A is a networking structure schematic diagram of a cloud cluster switch in an embodiment;

[0043] FIG. 7B is a schematic diagram of AI network health monitoring analysis and adjustment optimization in an embodiment;

[0044] FIG. 8A is a structural schematic diagram of a packet processing device in an embodiment;

[0045] FIG. 8B is a structural schematic diagram of a resource management device in an embodiment;

[0046] FIG. 9 is a hardware structure diagram of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0047] With the rapid development of the Internet, cloud services, big data and AI technology, the traffic of DC networks is growing rapidly, which puts a huge pressure on the network devices of DC networks. In particular, the development and application of AIGC lead to an exponential increase in computing and traffic. For example, the number of parameters of an AI large model reaches the trillion level, the size of the training data set reaches the TB level, and the interconnection of GPU (Graphics Processing Unit) of the order of ten thousand cards is needed to realize the pre-training of the model, resulting in a huge scale and complex connection of DC networks.

[0048] Referring to FIG. 1, a structural schematic diagram of a DC network is shown. The DC network includes a large number of Spine Switches and a large number of Leaf Switches, and FIG. 1 takes 32 Spine Switches and 64 Leaf Switches as an example. The servers (such as GPUs) are accessed through the Leaf Switches, and then the Spine Switches are connected to each Leaf Switch to realize data interaction between servers.

[0049] For the network devices (such as Spine Switches and Leaf Switches) of the DC network, the network devices include a main control board, an interface board and a network board. Due to the hardware structure limitation of the network devices, the main control board, the interface board and the network board cannot be expanded. Based on this, when the traffic grows a lot, the network devices cannot meet the forwarding demand of a large amount of traffic, resulting in problems such as increased delay and traffic packet loss.

[0050] To meet the forwarding requirement of a large amount of traffic, a large number of network devices, such as 32 Spine Switches and 64 Leaf Switches, are deployed. When a large number of network devices are deployed, traffic congestion occurs between different network devices, and the traffic congestion causes the computing capability of the server to decrease. For example, the traffic congestion between the Spine Switches causes the computing capability to decrease by 30%, thereby wasting the computing resources of the server.

[0051] In view of the above finding, the cloud cluster switch is proposed in the embodiment, the cloud cluster switch can be deployed in the DC network, and the cloud cluster switch replaces the networking architecture (Spine Switches and Leaf Switches) of the switch group of the DC network, to meet the development requirement of the Internet, cloud service, big data and AI technology.

[0052] The cloud cluster switch includes an NCC, an NCF and an NCP, the cloud cluster switch is a network device externally, the NCC, the NCF and the NCP in the cloud cluster switch are independent devices, and the NCC, the NCF and the NCP in the cloud cluster switch can be expanded. The cloud cluster switch can be expanded to meet the forwarding requirement of a large amount of traffic, reduce the forwarding delay, reduce the traffic packet loss and improve the service quality of the DC network.

[0053] Only a small number of cloud cluster switches need to be deployed to meet the forwarding requirement of a large amount of traffic, reduce the traffic congestion between the cloud cluster switches, improve the computing capability of the server and save the computing resources of the server.

[0054] The technical scheme of the cloud cluster switch in the embodiment can involve the following aspects:

[0055] First, physical decoupling of the network device form.

[0056] In the related art, the network device (such as a switch) mainly adopts a frame device, that is, the network device includes a main control board, an interface board and a network board, the main control board, the interface board and the network board are physically connected on a frame, the network device adopts a centralized architecture, and a large number of network devices are deployed in the DC network. However, when the network device of this type is deployed in the DC network of the AI scene, the following problems may exist: poor scalability: the expansion capability of the network device is limited, the network device needs to be increased to realize expansion, and the network may be complicated. In addition, the data flow model of the DC network of the AI scene is complex and changeable, a large number of network devices will cause network congestion, and the corresponding device and link bandwidth need to be quickly and timely adjusted. Poor fault redundancy: once the network device fails, the positioning and replacement are very complex, and the normal operation of the entire DC network may be affected.

[0057] In order to adapt to the AI scene, in this embodiment, a cloud cluster switch is introduced into the DC network, and the cloud cluster switch is connected with multiple box switches (i.e. independent devices) instead of the main control board, the interface board and the network board.

[0058] The cloud cluster switch can adopt a distributed architecture, and is connected with each box switch of the cloud cluster switch through a management network instead of a backplane bus, that is, each box switch of the cloud cluster switch is connected through the management network. Each box switch of the cloud cluster switch can be integrated into one device, and the cloud cluster switch has stronger expansibility.

[0059] The cloud cluster switch has good expansibility by increasing the device management function to realize the rapid expansion of the DC network. The cloud cluster switch has strong fault isolation capability: once a box switch fails, the fault can be isolated in time, and the operation of the entire DC network will not be affected. The cloud cluster switch has high flexibility: the structure of the cloud cluster switch is flexible, and the cloud cluster switch can be flexibly configured and deployed according to the service demand of the AI scene.

[0060] The cloud cluster switch can include an NCC (such as one or more NCCs), an NCF (such as one or more NCFs) and an NCP (such as multiple NCPs). Inside the cloud cluster switch, the NCC can be connected with the NCF, the NCC can be connected with each NCP respectively, and the NCF can be connected with each NCP respectively.

[0061] The NCC can be a box switch, that is, the function of the NCC is realized through the box switch, and the NCC corresponds to the main control board of the frame switch. The NCC is a cloud cluster controller, which is responsible for managing and controlling each member device in the cloud cluster switch and the network operation status. The NCC is responsible for the allocation and monitoring of network resources, and ensures the stability and safety of the operation of the entire cloud cluster switch.

[0062] The NCF can be a box switch, that is, the function of the NCF is realized through the box switch, and the NCF corresponds to the network board (switch network board) of the frame switch. The NCF is a cloud cluster forwarding device, which is responsible for message forwarding between NCPs, that is, forwarding the message of one NCP to another NCP.

[0063] The NCP can be a box switch, that is, the function of the NCP is realized through the box switch, and the NCP corresponds to the interface board of the frame switch. The NCP is a cloud cluster access device, which is responsible for accessing various types of servers, such as accessing GPU servers, and forwarding messages of the servers.

[0064] Second, the physical connection structure based on the cloud cluster switch.

[0065] Referring to Fig. 2, a schematic diagram of a topology of a cloud cluster switch is shown. The cloud cluster switch can include an NCC, an NCF, and a plurality of NCPs. Inside the cloud cluster switch, the NCC can be connected with the NCF, the NCC can be connected with each of the NCPs, and the NCF can be connected with each of the NCPs.

[0066] In Fig. 2, two NCCs (NCC1 and NCC2), two NCFs (NCF1 and NCF2), and four NCPs (NCP1, NCP2, NCP3, and NCP4) are taken as examples for illustration.

[0067] The NCC, the NCF, and the NCP are independent physical devices (e.g., box switches), and the NCC, the NCF, and the NCP are connected together through a MGT (Multi Gigabit Transceiver) network to form a cloud cluster switch, which presents as one physical device externally.

[0068] The MGT is a management network of the cloud cluster switch. In Fig. 2, three MGTs (MGT1, MGT2, and MGT3) are taken as examples. The MGT can be implemented through a three-layer switch or other devices, which is not limited. The MGT supports a routing protocol and a multicast protocol, and is physically connected with the NCC, the NCF, and the NCP.

[0069] Referring to Fig. 2, the NCC is connected with each of the NCPs through the MGT, the NCC is connected with each of the NCFs through the MGT (the connection relationship is not shown in Fig. 2), and the NCF is directly connected with each of the NCPs.

[0070] The NCP is responsible for accessing various types of servers. In Fig. 2, the NCP1 accesses a server, the NCP3 accesses a server, and the NCP4 accesses a Switch, which can access a plurality of servers.

[0071] Referring to Fig. 2, the NCC, the MGT, and the NCP form an LIPC network, and the NCPs are managed through an LIPC communication protocol. The NCF and the NCP form a CELL forwarding data network, which forwards data packets of each NCP. The NCP and the server form a service network, and the NCP can be connected with various servers or a lower-hanging switch.

[0072] In an example, a plurality of distributed box switches can be combined into one cloud cluster switch, that is, a plurality of box switches are combined into one switch (referred to as a cloud cluster switch) in a cloud cluster manner. The cloud cluster is a software virtualization technology, which realizes decoupling between an application and a physical device based on a containerization architecture. The cloud cluster is divided into two levels: a physical cluster at a physical device level and a container cluster at an application level.

[0073] For physical clusters at the physical device level, multiple physical devices (such as multiple box switches) can be connected together and, after configuration, can be virtualized into a single device (cloud cluster switch).

[0074] This virtualization technology can aggregate the hardware resources of multiple physical devices, enabling unified management and allocation of these resources, improving resource utilization, and reducing management complexity. This virtualization technology can also achieve hardware-level backup, enhancing the overall system reliability.

[0075] For application-level container clusters, containers running on physical devices can be logically connected together and, after configuration, virtualized into a single system. This virtualization technology can combine the software processing capabilities of multiple containers, enabling collaborative work, unified management, and uninterrupted maintenance of multiple containers.

[0076] Third, the composition of NCC, NCF and NCP and cloud cluster platform components.

[0077] For each member device of the cloud cluster switch (such as NCC, NCF, and NCP), the member device can use a specified operating system (such as Linux), and deploy the cloud cluster platform components (referred to as the cloud platform) on the specified operating system. Through container-based virtualization technology, the management functions and business functions of the member devices can be separated, allowing the management functions and business functions to run independently.

[0078] Referring to Figure 3A, this is a schematic diagram of the cloud cluster platform components. The cloud cluster platform components can include routing containers and application containers. Routing containers implement management functions, while application containers implement business functions. For example, a separate application container can be deployed for each business function.

[0079] Referring to Figure 3B, which is a schematic diagram of the cloud cluster platform components, the cloud cluster platform components can be divided into the following four components according to their roles: Cloud Platform Worker component, which runs on each member device; Cloud Platform Manager component, which runs on each cluster management node (i.e., the management node for all member devices); Cloud Platform Agent component, which runs inside the application container; and Cloud Platform Admin component, which runs on each member device and is used to receive and process configuration messages.

[0080] Cloud Platform Manager Component: running in the host operating system of each member device participating in cluster management, responsible for cloud platform HA (high availability) functions, establishing clusters, managing cluster members. The cloud platform Manager component is responsible for managing physical clusters, establishing physical clusters, maintaining physical clusters, managing cluster members, generating and updating cluster topology. Responsible for managing container clusters, can intelligently deploy application containers according to the distribution of hardware resources in the physical cluster, and elect the master container and slave container of the container cluster.

[0081] Cloud Platform Worker Component: running in the host operating system of each member device, responsible for managing member devices and container life cycles, periodically reporting physical resources and states of member devices, responding to scheduling instructions of the cloud platform Manager component, and creating and running containers according to instructions of the cloud platform Manager component.

[0082] Cloud Platform Admin Component: running on each member device, the cloud platform Admin component can receive and process configuration messages of the master application container. The cloud platform Admin component is responsible for managing device running modes, container description files, and sending requests for container deployment to the cloud platform Manager component.

[0083] Cloud Platform Agent Component: the cloud platform Agent component runs inside the container and is used to report the health status (such as normal state or abnormal state) of the business in the container and notify the business module of cluster and container events.

[0084] In one example, the cloud cluster platform components of various member devices (such as NCC, NCF, and NCP) can discover and communicate with each other through routing containers and management network connections. The role of the routing container is to connect all member devices through LIPC communication, used for topology discovery of various member devices, logically forming a whole device, which needs to support routing features (IGP) and multicast features.

[0085] For topology discovery and maintenance, based on routing features (IGP), neighbors can be established based on link-local addresses of member devices, link-local addresses are announced as device addresses, and link-local addresses are reported by route management for topology detection and management based on link-local addresses.

[0086] For IPC (Inter Process Communication) interaction of containers, application containers and routing containers can perform IPC interaction to meet the query management and LIPC transparent transmission of the application container itself.

[0087] Fourth, the startup and topology establishment process of each member device in the cloud cluster switch.

[0088] In one example, the connection topology of the cloud cluster switch (i.e., the physical cluster). The control packet of the cloud cluster switch is a three-layer IP packet, and the IP addresses of each member device in the cloud cluster switch are configured in the same network segment. The IP address information can be automatically obtained through DHCP (Dynamic Host Configuration Protocol), and the network segment is used to interact the control packet of the cloud cluster switch.

[0089] The connection topology of the cloud cluster switch can be completed by star connection, which has low requirements for the physical location of each member device and can be applied to networking with dispersed physical locations of member devices. When the star connection is used to complete the connection topology of the cloud cluster switch, intermediate devices can also be used to interconnect the member devices.

[0090] After each member device is powered on, the cloud platform Worker component, the cloud platform Admin component, and the cloud platform Agent component can be automatically run. The cloud platform Manager component is an optional component, i.e., the cloud platform Manager component can be run or not.

[0091] For example, when the member device needs to participate in the management of the physical cluster, the role of the member device can be configured as manager+worker. At this time, the member device will run the cloud platform Manager component.

[0092] In one example, the establishment of the cloud cluster switch (i.e., the physical cluster). When the cloud cluster switch is established, the identity of each member device is determined through the configuration file. For example, the member device can obtain the identity information of the member device from the configuration file, which indicates that the member device is an NCP, or the identity information indicates that the member device is an NCF, or the identity information indicates that the member device is an NCC.

[0093] When the cloud cluster switch is established, cluster planning also needs to be completed. For example, which member devices participate in the management of the cloud cluster switch, which member device serves as the Leader (i.e., selecting one NCC as the Leader from all NCCs) of the cloud cluster switch, the number of each member device, the network segment used for internal communication of the cloud cluster switch, etc. During the cluster planning process, the above contents can be written into the configuration file.

[0094] For each member device, the member device can start the cloud platform Worker component, the cloud platform Admin component and the cloud platform Manager component according to the configuration file. The cloud platform Manager component is elected as the Leader, the cloud platform Worker component and the Leader establish an internal communication channel, the cloud platform Worker component registers to the Leader and reports hardware resource information. The Leader notifies the cloud platform Worker component to start a container.

[0095] In one example, after each member device in the cloud cluster switch is started, the member device determines identity information of the member device in the cloud cluster switch. The identity information indicates that the member device is an NCP (in the subsequent process, the NCP can be distinguished as a first NCP and a second NCP), or the identity information indicates that the member device is an NCF, or the identity information indicates that the member device is an NCC.

[0096] After learning the identity information of each member device, an operation corresponding to the identity information can be performed. Referring to Fig. 4, a schematic diagram of a registration process of each member device is shown. The process can include:

[0097] In step 401, for each NCP (the identity information indicates that the member device is an NCP, such as a first NCP and a second NCP), the NCP establishes a first internal channel with the NCC. The NCP sends a first registration message to the NCC through the first internal channel, and the first registration message includes local port information of a first port of the NCP.

[0098] In one example, in the topology discovery and maintenance process, neighbors can be established based on IP addresses (such as link-local addresses) of the member devices, and therefore, the NCP can learn the IP address of the NCC. The first internal channel is established based on the IP address of the NCP and the IP address of the NCC, and the first internal channel includes the two IP addresses.

[0099] Obviously, based on the IP address of the NCC of the first internal channel, the NCP can send a message to the NCC through the first internal channel. Based on the IP address of the NCP of the first internal channel, the NCC can send a message to the NCP through the first internal channel. In this way, the NCP and the NCC can interact messages through the first internal channel.

[0100] The first internal channel is a transmission channel between the NCP and the NCC. Since it is a transmission channel between the member devices in the cloud cluster switch, the transmission channel is referred to as the first internal channel. The first internal channel is any type of channel, as long as the NCP and the NCC can interact messages, and no limitation is made thereto.

[0101] The NCP can include a plurality of ports (referred to as first ports of the NCP), and the NCP connects to the server through the first ports, i.e., the NCP can receive the message sent by the server through the first ports.

[0102] For each first port of the NCP, the local port information of the first port indicates which port on the NCP the first port is. For example, the local port information can be a port identifier, or other information that can uniquely indicate the first port, and the local port information is not limited in this regard.

[0103] The local port information of the first port is only valid for the NCP, i.e., the NCP knows which port on the NCP the first port is based on the local port information of the first port. The local port information of the first port is not valid for other member devices other than the NCP, because the local port information of different member devices can be the same, and if the local port information of the first port is valid for all member devices, it will cause the local port information to be repeated, and then it will not be possible to send data messages based on the local port information of the first port.

[0104] In one example, the NCP can send a first registration message to the NCC through the first internal channel, and the first registration message includes the local port information of all first ports of the NCP. For example, assuming that the NCP includes three first ports, the first registration message can include the local port information a1 of the first port 1, the local port information a2 of the first port 2, and the local port information a3 of the first port 3.

[0105] Step 402, for each NCF (the identity information indicates that the member device is the NCF), the NCF establishes a second internal channel with the NCC. The NCF sends a second registration message to the NCC through the second internal channel, and the second registration message includes the local port information of the second port and the information of the NCP connected to the second port.

[0106] In one example, the NCF can know the IP address of the NCC. The second internal channel is established based on the IP address of the NCF and the IP address of the NCC, and the second internal channel includes the two IP addresses.

[0107] Obviously, based on the IP address of the NCC of the second internal channel, the NCF can send a message to the NCC through the second internal channel. Based on the IP address of the NCF of the second internal channel, the NCC can send a message to the NCF through the second internal channel. In this way, the NCF and the NCC can interact messages through the second internal channel.

[0108] The second internal channel is a transmission channel between the NCF and the NCC. Since it is a transmission channel between member devices inside the cloud cluster switch, it is referred to as the second internal channel. The second internal channel is any type of channel as long as the NCF and the NCC can interact with messages, and is not limited in this regard.

[0109] The NCF can include a plurality of ports (referred to as second ports of the NCF), and the NCF connects the NCP through each second port, respectively, i.e., the NCF can receive a message sent by the NCP through the second port.

[0110] For each second port of the NCF, the local port information of the second port indicates which port on the NCF the second port is. For example, the local port information can be a port identifier, or other information that can uniquely indicate the second port, and the local port information is not limited in this regard. The local port information of the second port is only valid in the NCF, i.e., the NCF learns which port on the NCF the second port is based on the local port information. The local port information of the second port is not valid in other member devices other than the NCF.

[0111] In an example, during the topology discovery and maintenance process, for each second port, the NCF can learn the information of the NCP connected to the second port (such as the IP address of the NCP, etc.). For example, when the NCF establishes a neighbor with the NCP through the second port, it can learn the information of the NCP connected to the second port.

[0112] In an example, when the NCF sends the second registration message, the second registration message includes the local port information of the second port and the information of the NCP connected to the second port. For example, assuming that the NCF includes three second ports, the second registration message includes the local port information b1 of the second port 1 and the IP address of the NCP1 connected to the second port 1, the local port information b2 of the second port 2 and the IP address of the NCP2 connected to the second port 2, and the local port information b3 of the second port 3 and the IP address of the NCP3 connected to the second port 3.

[0113] Step 403, for each NCC (the identity information indicates that the member device is the NCC), the NCC receives the first registration message, configures the global port information for the first port of the NCP, and records the correspondence between the local port information of the first port and the global port information of the first port in the port mapping table.

[0114] In an example, for each local port information of the first port carried by the first registration message, the NCC can generate the global port information based on the local port information, i.e., configure the global port information for the first port, and the global port information is valid in all member devices, i.e., it is unique in all member devices.

[0115] For example, based on the local port information a1 of the first port 1 carried by the first registration message, the NCC can generate the global port information a1' for the first port 1. In generating the global port information a1', an algorithm can be used to operate on the local port information a1 to obtain the global port information a1'. For example, a random number is generated, and the random number is combined with the local port information a1 to obtain the global port information a1'.

[0116] Of course, the above is only an example of generating the global port information a1', and no limitation is made thereto. As long as the global port information a1' can be generated based on the local port information a1, and the global port information a1' is unique, it is acceptable.

[0117] Based on the local port information a2 of the first port 2 carried by the first registration message, the NCC can generate the global port information a2' for the first port 2. Based on the local port information a3 of the first port 3 carried by the first registration message, the NCC can generate the global port information a3' for the first port 3.

[0118] In one example, the NCC can record the correspondence between the local port information of the first port and the global port information of the first port in the port mapping table, see Table 1 for an example of the port mapping table.

[0119] Table 1

[0120] Step 404, for each NCC (the identity information indicates that the member device is an NCC), the NCC receives a second registration message, configures global port information for a second port of the NCF, and records the correspondence between the local port information of the second port, the global port information of the second port, and the information of the NCP in the port mapping table.

[0121] In one example, for each local port information of a second port carried by the second registration message, the NCC can generate global port information based on the local port information, i.e., configure the global port information for the second port, and the global port information is valid in all member devices, i.e., unique in all member devices.

[0122] For example, based on the local port information b1 of the second port 1 carried by the second registration message, the NCC can generate the global port information b1' for the second port 1. In generating the global port information b1', an algorithm can be used to operate on the local port information b1 to obtain the global port information b1'. For example, a random number is generated, and the random number is combined with the local port information b1 to obtain the global port information b1'.

[0123] Of course, the above is just an example of generating global port information b1', and is not limited thereto. As long as the global port information b1' can be generated based on the local port information b1, and the global port information b1' is unique, it is acceptable.

[0124] Based on the local port information b2 of the second port 2 carried by the second registration message, the NCC can generate global port information b2' for the second port 2. Based on the local port information b3 of the second port 3 carried by the second registration message, the NCC can generate global port information b3' for the second port 3.

[0125] In one example, the NCC can record in the port mapping table the correspondence between the local port information of the second port, the global port information of the second port, and the information of the NCP (such as the IP address of the NCP connected by the second port, obtained from the second registration message). See Table 2 for an example of the port mapping table.

[0126] Table 2

[0127] In step 405, for each NCP (the identity information indicates that the member device is an NCP, such as the first NCP and the second NCP), the NCP periodically sends a first heartbeat message to the NCC through the first internal channel, and the first heartbeat message includes resource information of the NCP, which indicates the remaining available resources of the NCP.

[0128] For example, the NCP can periodically count the resource information of the NCP. For example, the resource information can include but is not limited to one or more of memory resource information, CPU resource information, bandwidth resource information, and storage resource information. The memory resource information indicates the remaining available memory resources of the NCP, the CPU resource information indicates the remaining available CPU resources of the NCP, the bandwidth resource information indicates the remaining available bandwidth resources of the NCP, and the storage resource information indicates the remaining available storage resources of the NCP.

[0129] After counting the resource information each time, the NCP can send a first heartbeat message to the NCC through the first internal channel, and the first heartbeat message can include the resource information of the current counting period.

[0130] In step 406, for the NCC (the identity information indicates that the member device is an NCC), the NCC receives the first heartbeat message and obtains the resource information of the NCP from the first heartbeat message.

[0131] In one example, if the resource information of the NCP indicates that the remaining available resource of the NCP is less than a first threshold (which can be configured according to experience), the NCC displays a first alarm information, which is used to prompt to expand the NCP of the cloud cluster switch, i.e., to increase the number of NCPs in the cloud cluster switch. For example, a new NCP can be deployed in the cloud cluster switch to share the processing pressure of the existing NCP.

[0132] For example, if the resource information includes memory resource information, and the memory resource information indicates that the remaining available memory resource is less than the first threshold, the NCC displays the first alarm information. For another example, if the resource information includes CPU resource information, and the CPU resource information indicates that the remaining available CPU resource is less than the first threshold, the NCC displays the first alarm information. For yet another example, if the resource information includes bandwidth resource information, and the bandwidth resource information indicates that the remaining available bandwidth resource is less than the first threshold, the NCC displays the first alarm information.

[0133] In one example, if the resource information of the NCP indicates that the remaining available resource of the NCP is greater than a second threshold (which can be configured according to experience, and the second threshold can be greater than the first threshold), the NCC displays a second alarm information, which is used to prompt to shrink the NCP of the cloud cluster switch, i.e., to reduce the number of NCPs in the cloud cluster switch. For example, the existing NCPs can be offline from the cloud cluster switch.

[0134] For example, if the resource information includes memory resource information, and the memory resource information indicates that the remaining available memory resource is greater than the second threshold, the NCC displays the second alarm information. For another example, if the resource information includes CPU resource information, and the CPU resource information indicates that the remaining available CPU resource is greater than the second threshold, the NCC displays the second alarm information. For yet another example, if the resource information includes bandwidth resource information, and the bandwidth resource information indicates that the remaining available bandwidth resource is greater than the second threshold, the NCC displays the second alarm information.

[0135] In one example, if the resource information of the NCP indicates that the remaining available resource of the NCP is not less than the first threshold, and the remaining available resource is not greater than the second threshold, the NCC does not display alarm information, i.e., the NCP of the cloud cluster switch does not need to be expanded, and the NCP of the cloud cluster switch does not need to be shrunk.

[0136] Alternatively, the NCC can also display alarm information (prompt information), which is used to prompt not to expand the NCP of the cloud cluster switch, and not to shrink the NCP of the cloud cluster switch.

[0137] Step 407, for each NCF (identity information indicates that the member device is an NCF), the NCF periodically sends a second heartbeat message to the NCC through the second internal channel, and the second heartbeat message can include resource information of the NCF, which can represent the remaining available resources of the NCF.

[0138] For example, the NCF can periodically count the resource information of the NCF, which can include one or more of memory resource information, CPU resource information, bandwidth resource information, and storage resource information. After counting the resource information each time, the NCF can send a second heartbeat message to the NCC through the second internal channel, and the second heartbeat message can include the resource information of the current counting period.

[0139] Step 408, for the NCC (identity information indicates that the member device is an NCC), the NCC receives the second heartbeat message and obtains the resource information of the NCF from the second heartbeat message.

[0140] In one example, if the resource information of the NCF indicates that the remaining available resources of the NCF are less than a third threshold value (which can be configured according to experience), the NCC displays third alarm information, which is used to prompt the expansion of the NCF of the cloud cluster switch, i.e., increasing the number of NCFs in the cloud cluster switch. For example, a new NCF can be deployed in the cloud cluster switch to share the processing pressure of the existing NCF.

[0141] In one example, if the resource information of the NCF indicates that the remaining available resources of the NCF are greater than a fourth threshold value (which can be configured according to experience, and the fourth threshold value can be greater than the third threshold value), the NCC displays fourth alarm information, which is used to prompt the contraction of the NCF of the cloud cluster switch, i.e., reducing the number of NCFs in the cloud cluster switch. For example, an existing NCF can be offline from the cloud cluster switch.

[0142] In one example, if the resource information of the NCF indicates that the remaining available resources of the NCF are not less than the third threshold value and the remaining available resources are not greater than the fourth threshold value, the NCC does not display alarm information, i.e., the NCF of the cloud cluster switch does not need to be expanded, and the NCF of the cloud cluster switch does not need to be contracted.

[0143] Alternatively, the NCC can also display alarm information (prompt information), which is used to prompt that the NCF of the cloud cluster switch does not need to be expanded, and the NCF of the cloud cluster switch does not need to be contracted.

[0144] Fifth, packet processing process of each member device in the cloud cluster switch.

[0145] The embodiment of the present application provides a packet processing method, which can be applied to a member device in a cloud cluster switch, and the member device can be an NCC, an NCF or an NCP.

[0146] After each member device in the cloud cluster switch is started, the member device determines identity information of the member device in the cloud cluster switch. The identity information indicates that the member device is an NCP, or the identity information indicates that the member device is an NCF, or the identity information indicates that the member device is an NCC.

[0147] In one example, the NCPs can be divided into a first NCP and a second NCP. The first NCP is an NCP that receives a protocol packet (such as an ARP protocol packet) sent by a server, that is, the NCP that receives the protocol packet is referred to as the first NCP, and the remaining NCPs other than the first NCP are referred to as the second NCP.

[0148] In one example, for a member device in the cloud cluster switch, identity information of the member device in the cloud cluster switch can be determined. If the identity information indicates that the member device is the first NCP and a first protocol packet sent by a server is received through a first port, local port information of the first port is added to the first protocol packet to obtain a second protocol packet, and the second protocol packet is sent to the NCC.

[0149] If the identity information indicates that the member device is the NCC and the second protocol packet is received, global port information of the first port is determined based on the local port information of the first port. A first forwarding table item is established, and the first forwarding table item includes address information of the server in the first protocol packet and the global port information of the first port.

[0150] If the identity information indicates that the member device is the NCC, the first forwarding table item is sent to the first NCP, so that after the first NCP receives a data packet with the address information of the server as a destination address, the global port information of the first port is obtained based on the destination address by querying the first forwarding table item, and the data packet is sent through the first port.

[0151] As can be seen from the above technical solutions, in the embodiment of the present application, the cloud cluster switch is deployed in the DC network, and the cloud cluster switch includes the NCC, the NCF and the NCP. The NCC, the NCF and the NCP are independent devices (such as box-type switches), and the NCC, the NCF and the NCP can be expanded or contracted.

[0152] When the traffic increases greatly, the cloud cluster switch can also be expanded to meet the forwarding requirements of a large amount of traffic, thereby reducing the forwarding delay, reducing the traffic packet loss and improving the service quality of the DC network.

[0153] Referring to FIG. 5, a flowchart of a packet processing method is shown. The method can include the following steps.

[0154] In step 501, a first NCP receives a first protocol packet sent by a server through a first port, adds local port information of the first port to the first protocol packet to obtain a second protocol packet, and sends the second protocol packet to an NCC. The first NCP is a member device with an identity information indicating that it is an NCP, and the NCP receives the first protocol packet sent by the server. The first protocol packet can include address information of the server, such as an IP address of the server.

[0155] In one example, the first protocol packet is a protocol packet for triggering establishment of a table entry, and the first protocol packet can include address information of the server, such as an IP address of the server. For example, the first protocol packet can be an ARP protocol packet, or can be another type of protocol packet that can trigger establishment of a table entry.

[0156] When the first NCP receives the first protocol packet sent by the server through the first port 1, the first NCP can add local port information a1 of the first port 1 to the first protocol packet to obtain the second protocol packet.

[0157] In step 502, the first NCP sends the second protocol packet to the NCC.

[0158] In step 503, the NCC receives the second protocol packet, and determines global port information of the first port based on the local port information of the first port.

[0159] In one example, the NCC can obtain the local port information a1 of the first port 1 from the second protocol packet, and query a port mapping table shown in Table 1 by using the local port information a1 to obtain global port information a1' corresponding to the local port information a1. The global port information a1' is the global port information of the first port 1.

[0160] In one example, the NCC can obtain address information of the server, such as an IP address of the server, from the second protocol packet after receiving the second protocol packet.

[0161] In step 504, the NCC establishes a first forwarding table entry including the address information of the server and the global port information of the first port. The NCC sends the first forwarding table entry to the first NCP, and sends the first forwarding table entry to a second NCP. That is, the NCC sends the first forwarding table entry to each NCP.

[0162] In one example, the first forwarding table entry is used to guide the forwarding of the data packet, the matching item of the first forwarding table entry is the address information of the server, and the action item of the first forwarding table entry is the global port information of the first port. Of course, the first forwarding table entry can also include other contents, and the first forwarding table entry is not limited in this regard.

[0163] The NCC can send the first forwarding table entry to the first NCP, and the first NCP stores the first forwarding table entry. In the forwarding process of the data packet, the first NCP can forward the data packet based on the first forwarding table entry.

[0164] The NCC can send the first forwarding table entry to the second NCP, and the second NCP stores the first forwarding table entry. In the forwarding process of the data packet, the second NCP can forward the data packet based on the first forwarding table entry.

[0165] In step 505, the NCC establishes a second forwarding table entry, which includes the global port information of the first port and the global port information of the second port. The NCC sends the second forwarding table entry to the NCF.

[0166] The NCC determines the global port information of the second port of the NCF, and the NCF is connected to the first NCP through the second port, that is, the global port information of the second port of the NCF connected to the first NCP is determined.

[0167] For example, when the NCC receives the second protocol packet sent by the first NCP, it can also obtain the information of the first NCP, such as the IP address of NCP1, indicating that NCP1 sends the second protocol packet to the NCC.

[0168] After obtaining the information of the first NCP, the NCC queries the port mapping table through the information of the first NCP to obtain the local port information and the global port information of the second port. For example, when querying the port mapping table shown in Table 2 through the IP address of NCP1, the global port information b1' of the second port can be obtained.

[0169] After obtaining the global port information of the first port and the global port information of the second port, the NCC can establish a second forwarding table entry. The second forwarding table entry is used to guide the forwarding of the data packet, the matching item of the second forwarding table entry is the global port information of the first port, and the action item of the second forwarding table entry is the global port information of the second port. Of course, the second forwarding table entry can also include other contents, and the second forwarding table entry is not limited in this regard.

[0170] The NCC can send the second forwarding table entry to each NCF, and the NCF stores the second forwarding table entry. In the forwarding process of the data packet, the NCF can forward the data packet based on the second forwarding table entry.

[0171] In summary, each NCP stores the first forwarding table item, and each NCF stores the second forwarding table item. The following describes the data packet forwarding process based on the first forwarding table item and the first forwarding table item.

[0172] In step 506, the second NCP receives the data packet, and the destination address of the data packet is the address information of the server. The second NCP queries the first forwarding table item based on the destination address to obtain the global port information of the first port. If it is determined based on the global port information that the egress port corresponding to the data packet is not located in the second NCP, the second NCP encapsulates the global port information of the first port for the data packet and sends the encapsulated data packet to the NCF.

[0173] Since the first forwarding table item includes the address information of the server and the global port information of the first port, the destination address of the data packet is the address information of the server. Therefore, the second NCP can obtain the global port information corresponding to the destination address when querying the first forwarding table item based on the destination address of the data packet.

[0174] When generating the global port information, the NCC performs an operation on the local port information to obtain the global port information by using an algorithm, such as combining a random number with the local port information to obtain the global port information. After obtaining the global port information corresponding to the destination address, the second NCP can also perform an inverse operation on the global port information to obtain the local port information, such as removing the random number in the global port information to obtain the local port information.

[0175] After obtaining the local port information, the second NCP can determine whether the first port corresponding to the local port information (that is, the egress port corresponding to the data packet) is located in the second NCP. If not, the second NCP encapsulates the global port information of the first port for the data packet and sends the encapsulated data packet to the NCF.

[0176] In step 507, the NCF receives the data packet. If the data packet includes the global port information of the first port, the NCF queries the second forwarding table item to obtain the global port information of the second port and sends the data packet through the second port.

[0177] Since the second NCP encapsulates the global port information of the first port for the data packet and sends the encapsulated data packet to the NCF, the NCF can obtain the global port information of the first port from the data packet after receiving the data packet and query the second forwarding table item based on the global port information of the first port.

[0178] Since the second forwarding table item includes the global port information of the first port and the global port information of the second port, the NCF can obtain the global port information of the second port when querying the second forwarding table item.

[0179] When generating the global port information, the NCC performs operation on the local port information by using an algorithm to obtain the global port information, for example, combining a random number with the local port information to obtain the global port information. After obtaining the global port information of the second port, the NCF can also perform reverse operation on the global port information to obtain the local port information of the second port, for example, removing the random number in the global port information to obtain the local port information.

[0180] After obtaining the local port information of the second port, the NCF can determine the second port corresponding to the local port information, that is, which second port of the NCF the local port information corresponds to. On this basis, the NCF can send the data packet through the second port, that is, send the data packet to the first NCP.

[0181] Before the NCF sends the data packet through the second port, the global port information of the first port can also be removed, that is, the first port global port information encapsulated by the second NCP for the data packet.

[0182] In step 508, the first NCP receives the data packet, and the destination address of the data packet is the address information of the server. The first NCP queries the first forwarding table item based on the destination address to obtain the global port information of the first port. If it is determined based on the global port information that the egress port corresponding to the data packet is located in the NCP, the data packet is sent through the first port, that is, the first NCP sends the data packet to the server through the first port.

[0183] Since the first forwarding table item includes the address information of the server and the global port information of the first port, the destination address of the data packet is the address information of the server. Therefore, when the first NCP queries the first forwarding table item based on the destination address of the data packet, the global port information corresponding to the destination address can be obtained.

[0184] When generating the global port information, the NCC performs operation on the local port information by using an algorithm to obtain the global port information, for example, combining a random number with the local port information to obtain the global port information. After obtaining the global port information corresponding to the destination address, the first NCP can also perform reverse operation on the global port information to obtain the local port information, for example, removing the random number in the global port information to obtain the local port information.

[0185] After obtaining the local port information, the first NCP can determine whether the first port corresponding to the local port information (i.e., the out port corresponding to the data packet) is located in the first NCP. If yes, the first NCP sends the data packet through the first port, thereby sending the data packet to the server.

[0186] For the data packet received by the first NCP, the data packet can be originated from the second NCP, i.e., the second NCP sends the data packet to the NCF, and the NCF sends the data packet to the first NCP. The data packet can also be originated from another server connected to the first NCP, i.e., the server directly sends the data packet to the first NCP, in which case the first NCP directly forwards the data packet through the first forwarding table item.

[0187] Up to now, the forwarding process of the data packet is completed, and the data packet is successfully sent to the server.

[0188] In one example, the NCP includes an application container and a routing container, the NCC includes an application container and a routing container, and the NCF includes an application container and a routing container. The steps 501-505 are described as follows.

[0189] For the steps 501 and 502, for the first NCP, the application container of the first NCP is configured to receive the first protocol packet and send the first protocol packet to the routing container of the first NCP.

[0190] The routing container of the first NCP is configured to add the local port information of the first port to the first protocol packet to obtain a second protocol packet and send the second protocol packet to the routing container of the NCC.

[0191] For the steps 503 and 504, for the NCC, the routing container of the NCC is configured to receive the second protocol packet and send the second protocol packet to the application container of the NCC.

[0192] The application container of the NCC is configured to determine the global port information of the first port based on the local port information of the first port, establish a first forwarding table item, and send the first forwarding table item to the application container of the first NCP.

[0193] The application container of the NCC is configured to send the first forwarding table item to the application container of the second NCP.

[0194] The application container of the NCC is configured to determine the global port information of the second port of the NCF, establish a second forwarding table item, and the application container of the NCC is configured to send the second forwarding table item to the application container of the NCF.

[0195] From the above technical solutions, in the embodiment of the present application, the DC network deploys a cloud cluster switch, and the cloud cluster switch includes an NCC, an NCF and an NCP. The NCC, the NCF and the NCP are independent devices (such as box switches), and the NCC, the NCF and the NCP can be expanded or contracted.

[0196] When the traffic increases greatly, the cloud cluster switch can also meet the forwarding requirement of a large amount of traffic through expansion, thereby reducing the forwarding delay, reducing the traffic packet loss and improving the service quality of the DC network.

[0197] Sixth, the joining of a member device in the cloud cluster switch or the leaving of the member device.

[0198] When the first member device (the member device to be added to the cloud cluster switch is referred to as the first member device) joins the cloud cluster switch, the NCC establishes a third internal channel with the first member device, and receives a third registration message sent by the first member device through the third internal channel.

[0199] For example, the first member device establishes the third internal channel with the NCC, and sends the third registration message to the NCC through the third internal channel. In this way, the NCC can establish the third internal channel with the first member device, and receive the third registration message sent by the first member device through the third internal channel.

[0200] Based on the third registration message, the first member device (such as the NCP or the NCF) is registered on the NCC. The registration process of the first member device can be referred to FIG. 4, and steps 401-404.

[0201] In addition to the registration process, after receiving the third registration message, the NCC can also perform the following operations:

[0202] If the first member device is a second NCP (the identity information of the first member device indicates the NCP, and the NCP newly added to the cloud cluster switch is the second NCP) and the third registration message is received, all the first forwarding table entries are sent to the second NCP. In this way, the second NCP can be guided to forward the data packet based on the first forwarding table entries. The forwarding process of the data packet can be referred to the above embodiment.

[0203] If the first member device is an NCF (the identity information of the first member device indicates the NCF) and the third registration message is received, all the second forwarding table entries are sent to the NCF. In this way, the NCF can be guided to forward the data packet based on the second forwarding table entries. The forwarding process can be referred to the above embodiment.

[0204] In one example, the identity information of a newly added member device of a cloud cluster switch (physical cluster) can also be determined by a configuration file. The joining process of a member device is described by taking the joining process of Device B (i.e., a first member device) to the physical cluster Device A (i.e., an NCC) as an example.

[0205] Device B starts the cloud platform Admin component and the cloud platform Worker component according to the configuration file. The cloud platform Worker component automatically starts an application container, and the cloud platform Worker component automatically registers with the Leader and starts a cluster joining timer. Device A is the Leader in the physical cluster, and the cloud platform Worker component is replied to join successfully. The Leader issues a command to create a container, and the Leader schedules Device B according to the resource information reported by each member device, creates and starts a container on Device B. After the container of Device B is successfully started, the cloud platform Worker component reports the container information to the Leader. The Leader synchronizes the forwarding table item information of other containers in the current cloud cluster to the cloud platform Worker component of Device B.

[0206] When a second member device (a member device to be left from a cloud cluster switch is referred to as a second member device) leaves the cloud cluster switch, the NCC performs the following processing based on the identity of the second member device.

[0207] If the second member device is a second NCP, the second member device directly leaves the cloud cluster switch when the second member device leaves the cloud cluster switch, and the update process of the first forwarding table item and the second forwarding table item is not triggered.

[0208] If the second member device is an NCF, the second member device directly leaves the cloud cluster switch when the second member device leaves the cloud cluster switch, and the update process of the first forwarding table item and the second forwarding table item is not triggered.

[0209] If the second member device is a first NCP, the NCC can also obtain a third forwarding table item and a fourth forwarding table item corresponding to the second member device, the third forwarding table item is a first forwarding table item triggered by the first NCP, and the fourth forwarding table item is a second forwarding table item triggered by the first NCP.

[0210] The third forwarding table item includes address information of a server and global port information of a port on the second member device, and the fourth forwarding table item includes global port information of the port on the second member device and global port information of a port on the NCF.

[0211] Based on this, the NCC can send a first deletion instruction to all second NCPs, which is used to make each second NCP delete the third forwarding table item corresponding to the second member device. In this way, the second NCP can delete the third forwarding table item, and thus will not send data packets to the second member device based on the third forwarding table item, thereby avoiding data packet transmission anomalies caused by the second member device leaving.

[0212] In addition, the NCC can send a second deletion instruction to all NCFs, which is used to make each NCF delete the fourth forwarding table item corresponding to the second member device. In this way, the NCF can delete the fourth forwarding table item, and thus will not send data packets to the second member device based on the second forwarding table item.

[0213] In one example, when the second member device leaves the cloud cluster switch, it can be divided into two cases:

[0214] The second member device actively leaves the cloud cluster switch. For example, the administrator removes the second member device from the cloud cluster switch (physical cluster). Before the second member device leaves the cloud cluster switch, it can send a leave cluster packet to the NCC (i.e., Leader), and the NCC will reply with a leave cluster response packet.

[0215] After the NCC receives the leave cluster packet, it determines that the second member device has left the cloud cluster switch. If the second member device is a second NCP or an NCF, the NCC deletes the second member device from the physical cluster device list and the physical cluster topology, and synchronizes the updated physical cluster information and cluster topology to each member device. If the second member device is a first NCP, the NCC deletes the second member device from the physical cluster device list and the physical cluster topology, and synchronizes the updated physical cluster information and cluster topology to each member device. In addition, the NCC notifies all second NCPs to delete the third forwarding table item corresponding to the second member device, and the NCC notifies all NCFs to delete the fourth forwarding table item corresponding to the second member device.

[0216] When the administrator removes the second member device from the cloud cluster switch, the undo join-cluster command is executed under the cloud-cluster member view to remove the second member device from the cloud cluster switch.

[0217] The second member device is passively left from the cloud cluster switch. When the route between the second member device and the NCC (Leader) is unreachable, the NCC cannot receive the heartbeat message sent by the second member device, and the NCC determines that the second member device leaves the cloud cluster switch. The NCC deletes the second member device from the physical cluster device list and the physical cluster topology, and synchronizes the updated physical cluster information and the cluster topology to each member device. If the second member device is the first NCP, the NCC instructs all second NCPs to delete the third forwarding table item corresponding to the second member device, and the NCC instructs all NCFs to delete the fourth forwarding table item corresponding to the second member device.

[0218] From the above technical solutions, in the embodiments of the present application, the cloud cluster switch is deployed in the DC network, and the cloud cluster switch includes an NCC, an NCF and an NCP. The NCC, the NCF and the NCP are independent devices, which can expand the NCC, the NCF and the NCP in the cloud cluster switch. The network bandwidth utilization can be improved and the fault recovery time can be reduced. The optimal choice is provided for the connection of different scales of computing power. The AI computing power scene is designed, which has strong ecological decoupling ability and excellent computing power network performance. The best solution for heterogeneous GPU network interconnection is provided, so that the network can match the business flow model of the AI intelligent computing center.

[0219] Seventh, resource management of the cloud cluster switch.

[0220] In the embodiments of the present application, a resource management method is provided, which can be applied to a management device. Referring to FIG. 6, a flowchart of the resource management method is shown. The method can include the following steps:

[0221] In step 601, target resource data of the cloud cluster switch is obtained. The target resource data includes one or more of network element level resource data, network level resource data, application level resource data and service resource data.

[0222] In step 602, initial demand resource information of a user is obtained. The initial demand resource information represents an expected occupied resource, i.e., the user hopes to occupy a resource size matching the expected occupied resource.

[0223] In step 603, target demand resource information of the user is determined based on the target resource data and the initial demand resource information. The target demand resource information represents the actual allocated resource for the user.

[0224] In step 604, the target demand resource information is sent to the cloud cluster switch, so that the cloud cluster switch allocates available resources for the user according to the target demand resource information when receiving a data packet for the user, and processes the data packet based on the available resources.

[0225] In one example, the network-level resource data can include, but is not limited to, resource data of the cloud cluster switch, the resource data of the cloud cluster switch including network overall utilization, network capacity and network health of the cloud cluster switch. The network element-level resource data can include, but is not limited to, resource data of each member device in the cloud cluster switch, the member device being an NCC, an NCF or an NCP; the resource data of the member device including physical resource utilization, logical resource utilization and health of the member device.

[0226] In one example, the application-level resource data can include, but is not limited to, network quality when the cloud cluster switch forwards data packets, the network quality including one or more of packet loss rate, delay and jitter.

[0227] In one example, the application-level resource data corresponding to the cloud cluster switch is obtained, and specifically includes:

[0228] An RPC protocol packet sent by an RPC module of a member device in the cloud cluster switch is received, the RPC protocol packet can include monitoring indicators counted by the member device. The member device can be an NCC, an NCF or an NCP. The member device includes an SQA module, an eMDI module and an RPC module. The SQA module sends a target traffic feature of a specified type to the eMDI module, the specified type including an audio type and / or a video type; if the eMDI module receives a data packet matching the target traffic feature, the eMDI module counts monitoring indicators of the data packet and sends the monitoring indicators to the SQA module; the SQA module sends the monitoring indicators to the RPC module, and the RPC module generates an RPC protocol packet including the monitoring indicators.

[0229] The monitoring indicators counted by the member device are obtained from the RPC protocol packet.

[0230] Based on the monitoring indicators counted by each member device, application-level resource data is determined.

[0231] In one example, the service resource data can include, but is not limited to, resource data of a server accessing the cloud cluster switch; if the server is a GPU, the service resource data includes operation performance of the GPU and / or memory bandwidth of the GPU. The operation performance represents theoretical calculation capability of the GPU when processing data of different precisions. The memory bandwidth represents data transmission speed provided by the GPU when processing large-scale data.

[0232] In one example, determining the target demand resource information of the user includes: if the initial demand resource information includes an expected occupied resource for a cloud cluster switch, and the target resource data indicates that the available resource of the cloud cluster switch is greater than a first resource threshold, then the target demand resource information is greater than the initial demand resource information, so that the available resource allocated to the user is greater than the expected occupied resource. Alternatively, if the target resource data indicates that the available resource of the cloud cluster switch is less than a second resource threshold, then the target demand resource information is less than the initial demand resource information, so that the available resource allocated to the user is less than the expected occupied resource.

[0233] In one example, determining the target demand resource information of the user includes: if the initial demand resource information includes an expected occupied resource for a server connected to a cloud cluster switch, and the target resource data indicates that the available resource of the server is greater than a third resource threshold, then the target demand resource information is greater than the initial demand resource information, so that the available resource allocated to the user is greater than the expected occupied resource. Alternatively, if the target resource data indicates that the available resource of the server is less than a fourth resource threshold, then the target demand resource information is less than the initial demand resource information, so that the available resource allocated to the user is less than the expected occupied resource.

[0234] Eighth, AI network health monitoring analysis and adjustment optimization.

[0235] Based on the cloud cluster switch, the AI intelligent computing center can complete complex network construction by using one or more cloud cluster switches according to the size, and can manage the resources of the cloud cluster switch through a management device. For example, the cloud platform Manager component of the cloud cluster switch provides a management interface, and the management device calls the management interface to manage the resources of the cloud cluster switch.

[0236] Referring to FIG. 7A, it is a networking structure diagram of a cloud cluster switch. Here, cloud cluster switch 1 and cloud cluster switch 2 are taken as examples. Cloud cluster switch 1 includes cloud platform Manager 1 component, and cloud cluster switch 2 includes cloud platform Manager 2 component. The management device includes a controller cluster, an intelligent analysis module cluster, and a TCP (Transmission Control Protocol) collector.

[0237] In the resource management process, the TCP collector is configured to acquire data information in the running process of the cloud cluster switch. The intelligent analysis module cluster is configured to analyze target resource data based on the data information, the target resource data including network element level resource data, network level resource data, application level resource data and service resource data. Based on the target resource data and initial demand resource information, target demand resource information is determined. The controller cluster is configured to send the target demand resource information to the cloud cluster switch, so that the cloud cluster switch allocates available resources for the user according to the target demand resource information when receiving a data packet for the user.

[0238] In one example, the entire network is adapted to the traffic model of the AI intelligence center, as shown in FIG. 7B, which is a schematic diagram of AI network health monitoring, analysis and adjustment optimization. The process can include sensing, collecting, analyzing, optimizing and adjusting, which can make the network better match the intelligence center business.

[0239] 1. Sensing module and function.

[0240] The cloud platform Manager component interfaces various controller clusters of the AI intelligent computing center through a northbound network API (Application Programming Interface). The controller cluster can include a configuration management controller, a computing platform controller, a storage platform controller and a service operation controller.

[0241] The cloud platform Manager component interfaces the configuration management controller, which saves the role definition of each member device of the cloud cluster switch and the corresponding startup configuration template. The cloud platform Manager component acquires the startup configuration template for each member device through TFTP (Trivial File Transfer Protocol) and acquires the IP address through DHCP, thereby completing the startup construction of the cloud cluster switch.

[0242] The cloud platform Manager component interfaces the service operation controller and the computing platform controller, acquires the importance of tenants and their services, the computing resources (virtual machines and GPUs) corresponding to the tenants, and acquires the access port and network service quality policy template corresponding to the tenants, and then sends them to the corresponding port to complete the deployment of the service.

[0243] 2. Collection module and function, analysis module and function.

[0244] The collection module is configured to collect data information in the running process of the cloud cluster switch. The analysis module is configured to analyze the data information to obtain target resource data of the cloud cluster switch, which can include network element level resource data, network level resource data, and application level resource data. For example, the collection module can be deployed in a TCP collector, and the analysis module can be deployed in an intelligent analysis module cluster.

[0245] For the network element level resource data, the network element level resource data is resource data of each member device in the cloud cluster switch. The resource data of the member device can include physical resource utilization, logical resource utilization, and health degree of the member device. The member device can be an NCC, an NCF, or an NCP.

[0246] For example, the network element level resource data can include, but is not limited to, physical resource utilization, logical resource utilization, and health degree of each NCC, physical resource utilization, logical resource utilization, and health degree of each NCF, and physical resource utilization, logical resource utilization, and health degree of each NCP.

[0247] The physical resource utilization can be utilization of physical resources such as CPU, memory, and bandwidth. The logical resource utilization can be utilization of logical resources such as table entries (e.g., ARP table entries, forwarding table entries, etc.). The health degree represents the health degree of the member device. For example, when the CPU usage, bandwidth usage, etc. is large, the health degree is small, and when the CPU usage, bandwidth usage, etc. is small, the health degree is large.

[0248] For example, for each member device of the cloud cluster switch, the member device can support at least one of the following detection technologies: one-key self-check, one-key diagnosis, KPI (Key Performance Indicator), GOLD (Generic OnLine Diagnostics), and EAA (Embedded Automation Architecture).

[0249] Based on this, based on the detection technologies, the collection module can collect data information in the running process of the member device, and the analysis module can analyze the data information to obtain resource data of the member device.

[0250] Based on the resource data of each member device, the network element level resource data of the cloud cluster switch can be obtained, and the network element level resource data includes resource data of each member device in the cloud cluster switch.

[0251] For network-level resource data, the network-level resource data is resource data of the cloud cluster switch, and the resource data of the cloud cluster switch includes network overall utilization, network capacity and network health degree of the cloud cluster switch. Obviously, the network-level resource data is resource data for the overall network of the cloud cluster switch.

[0252] The network overall utilization refers to the degree of effective use of network resources, and is an important indicator for measuring network performance. The network overall utilization is divided into channel utilization and network utilization. The channel utilization refers to the ratio of data passing time to (with + without) data passing time. The network utilization is a weighted average value of the channel utilization, reflecting the comprehensive utilization of the entire network. The calculation of the network overall utilization involves multiple factors, such as communication capacity of the channel, data transmission speed, communication volume in the network at the same time, etc.

[0253] The network capacity is an important indicator for describing network performance, and is the amount of data that can be theoretically transmitted. The network capacity is related to various factors, such as bandwidth, load, throughput and maximum forwarding rate, etc.

[0254] The network health degree represents the health degree of the network of the cloud cluster switch.

[0255] For example, the cloud cluster switch can support at least one of the following detection technologies (i.e., network-level detection technology): NQA (Network Quality Analyzer), TWAMP (Two Way Active Measurement Protocol), TWAMP Light, iNQA (Intelligent Network Quality Analyzer), iFIT (In-situ Flow Information Telemetry), INT (In-Band Network Telemetry) and MOD (Mirror On Drop), without limitation to the detection technology.

[0256] The above detection technology is a network intelligent operation and maintenance technology, which is used for testing link state, evaluating network device performance and service quality, realizing detection of network performance, and effectively diagnosing and positioning faults when the network fails. Based on these detection technologies, the collection module can collect data information (such as network quality data such as packet loss rate, delay, delay jitter, etc.) in the running process of the cloud cluster switch, and the analysis module can analyze the data information to obtain the resource data of the cloud cluster switch, i.e., the network-level resource data.

[0257] For the application-level resource data, the application-level resource data can be network quality when the data packet is forwarded by the cloud cluster switch, and the network quality can include, but is not limited to, packet loss rate, delay and jitter.

[0258] For example, part of the member devices or all of the member devices of the cloud cluster switch are selected as the candidate member devices. The data information (such as packet loss rate, delay, jitter, etc.) in the running process of the candidate member devices is collected by the collection module, the analysis module can analyze the data information, obtain the network quality of the candidate member devices, and determine the application-level resource data based on the network quality of all the candidate member devices.

[0259] For example, the average of the packet loss rate of all the candidate member devices, the average of the delay of all the candidate member devices, and the average of the jitter of all the candidate member devices are taken as the application-level resource data.

[0260] For example, the cloud cluster switch can support at least one of the following detection technologies: NQA, TWAMP, TWAMP Light, iNQA, iFIT, INT and MOD, based on which the collection module can collect the data information (such as packet loss rate, delay, jitter, etc.) in the running process of the candidate member devices.

[0261] The analysis module can analyze the data information to obtain the network quality of the candidate member devices, and determine the application-level resource data based on the network quality of all the candidate member devices.

[0262] In one example, the member device enabled with eMDI (Enhanced Media Delivery Index) function can be taken as the candidate member device, and the member device without eMDI function is not taken as the candidate member device. The candidate member device can include SQA (Service Quality Analysis) module, eMDI module and RPC (Remote Procedure Call) module. Based on this, the following steps can be used to obtain the application-level resource data:

[0263] Step S11, the SQA module of the candidate member device sends a target traffic feature of a specified type to the eMDI module of the candidate member device, and the specified type includes audio type and / or video type.

[0264] In one example, the cloud platform Manager component can enable the eMDI function in the application container of the member device, and the member device with the enabled eMDI function is a candidate member device. On the candidate member device, the SQA module identifies the target traffic feature of the audio type and / or the target traffic feature of the video type, and sends the target traffic feature of the audio type and / or the target traffic feature of the video type to the eMDI module.

[0265] In step S12, the eMDI module monitors the data packet, and if a data packet matching the target traffic feature is received, the monitoring index of the data packet is counted, and the monitoring index is sent to the SQA module.

[0266] In one example, for the audio and video services of the AI intelligent computing center, the eMDI is a fault definition and health diagnosis technology designed for audio and video services (such as VoIP and IPTV). By enabling the eMDI function on the candidate member device, the eMDI module can monitor the data packet of the audio and video service in real time.

[0267] When the candidate member device receives a data packet each time, the eMDI module can determine whether the traffic feature of the data packet matches the target traffic feature. If yes, it means that the data packet is the data packet of the audio and video service, and the eMDI module needs to count the monitoring index of the data packet. If no, it means that the data packet is not the data packet of the audio and video service, and the monitoring index of the data packet is not counted.

[0268] When the eMDI module counts the monitoring index of the data packet, the monitoring index can be the data information in the running process of the candidate member device, such as the packet loss rate information, the time delay information, the jitter information, etc. After obtaining the monitoring index, the eMDI module can also send the monitoring index to the SQA module.

[0269] In step S13, the SQA module sends the monitoring index of the candidate member device to the RPC module.

[0270] In step S14, the RPC module encapsulates the monitoring index of the candidate member device into the RPC protocol packet, and sends the RPC protocol packet to the management device (such as the collection module of the management device).

[0271] In step S15, the management device receives the RPC protocol packet, and obtains the monitoring index counted by the candidate member device from the RPC protocol packet. The management device determines the application-level resource data based on the monitoring index counted by each candidate member device.

[0272] In one example, after receiving the RPC protocol packet, the collection module obtains the monitoring indexes of the candidate member device from the RPC protocol packet, and sends the monitoring indexes of the candidate member device to the analysis module. The analysis module analyzes the monitoring indexes of the candidate member device to obtain the network quality (such as packet loss rate, time delay, jitter, etc.) of the candidate member device, and determines the application-level resource data based on the network quality of all candidate member devices. For example, the average value of the packet loss rate of all candidate member devices, the average value of the time delay of all candidate member devices, and the average value of the jitter of all candidate member devices are taken as the application-level resource data.

[0273] In summary, the eMDI module, the SQA module, the RPC module, and the management device are used together to obtain the application-level resource data for the audio and video task, and thus the real-time application service condition is obtained.

[0274] If the application service response becomes slow, it is determined whether the network or the computing resource is affected based on the application-level resource data, and thus the corresponding optimization adjustment is adopted. Based on the application-level resource data, the network condition can be understood, the location of the network fault can be determined, the network deployment can be optimized, and the audio and video service requirement can be met.

[0275] In one example, in addition to the network element-level resource data, the network-level resource data, and the application-level resource data, the target resource data can further include service resource data, and the service resource data is the resource data of the server connected to the cloud cluster switch, which is used to indicate which resources the server has.

[0276] For example, if the server is a GPU, the service resource data can include but is not limited to the computing performance of the GPU and / or the memory bandwidth of the GPU, and of course, the service resource data is not limited to this.

[0277] The computing performance indicates the theoretical calculation capability of the GPU when processing different precision data. For example, the theoretical floating point operation performance of the GPU can be 9.7, 19.5, 156 TFLOPS (floating point operation times per second), etc., which represent the theoretical calculation capability when processing different precision data.

[0278] The memory bandwidth indicates the data transmission speed provided by the GPU when processing large-scale data. For example, the GPU is equipped with 40 GB of video memory, and the GPU has excellent memory bandwidth, such as the memory bandwidth can reach 1.6 TB / s, which means that high-speed data transmission can be provided when processing large-scale data.

[0279] In one example, the analysis module can obtain the work indicators of each GPU (server). If the work indicators are significantly lower than the self-capacity, the analysis of the existing reasons can be two, one is that the business is not busy, and the other is that the cooperation between GPUs, network communication congestion leads to the decline of computing power.

[0280] 3. The optimization module and functions, the adjustment module and functions.

[0281] The optimization module can determine the target demand resource information based on the target resource data and the initial demand resource information. The initial demand resource information represents the expected occupied resources, which can be provided by the user himself. The target demand resource information represents the actual allocated resources for the user.

[0282] In one example, if the initial demand resource information is the expected occupied resources for the cloud cluster switch, such as the user expecting to occupy 500M bandwidth resources. Then: if the target resource data indicates that the available resources of the cloud cluster switch are greater than the first resource threshold, the target demand resource information can be greater than the initial demand resource information, such as actually allocating 600M bandwidth resources for the user. The target demand resource information can also be equal to the initial demand resource information, such as actually allocating 500M bandwidth resources for the user. Or, if the target resource data indicates that the available resources of the cloud cluster switch are less than the second resource threshold, the target demand resource information can be less than the initial demand resource information, such as actually allocating 400M bandwidth resources for the user.

[0283] For example, the target resource data includes the physical resource utilization, logical resource utilization and health degree of each member device, the network overall utilization, network capacity and network health degree of the cloud cluster switch, and the network quality (packet loss rate, delay and jitter) when the cloud cluster switch forwards data packets.

[0284] If the physical resource utilization, logical resource utilization and health degree of the member device indicate that the available resources of the member device are more (such as more available bandwidth resources), the target resource data indicates that the available resources of the cloud cluster switch are greater than the first resource threshold. If the physical resource utilization, logical resource utilization and health degree of the member device indicate that the available resources of the member device are less (such as less available bandwidth resources), the target resource data indicates that the available resources of the cloud cluster switch are less than the second resource threshold.

[0285] If the network overall utilization, network capacity and network health of the cloud cluster switch represent that the network condition of the cloud cluster switch is better (i.e. more available bandwidth resources), the target resource data represents that the available resources of the cloud cluster switch are greater than the first resource threshold. If the network overall utilization, network capacity and network health of the cloud cluster switch represent that the network condition of the cloud cluster switch is worse, the target resource data represents that the available resources of the cloud cluster switch are less than the second resource threshold.

[0286] If the network quality (packet loss rate, latency and jitter) is better, the target resource data represents that the available resources of the cloud cluster switch are greater than the first resource threshold. If the network quality (packet loss rate, latency and jitter) is worse, the target resource data represents that the available resources of the cloud cluster switch are less than the second resource threshold.

[0287] In one example, if the initial demand resource information is the expected occupied resource of the server accessing the cloud cluster switch, for example, the user expects to occupy 10G of computing resources. Then: if the target resource data represents that the available resources of the server are greater than the third resource threshold, the target demand resource information can be greater than the initial demand resource information, for example, 15G of computing resources are actually allocated to the user. The target demand resource information can also be equal to the initial demand resource information, for example, 10G of computing resources are actually allocated to the user. Or, if the target resource data represents that the available resources of the server are less than the fourth resource threshold, the target demand resource information can be less than the initial demand resource information, for example, 8G of computing resources are actually allocated to the user.

[0288] For example, the target resource data includes service resource data, and the service resource data can include the available resources of the server (such as the size of the remaining computing resources), based on which it can be known that the available resources of the server are greater than the third resource threshold or less than the fourth resource threshold.

[0289] After obtaining the target demand resource information, the adjusting module sends the target demand resource information to the cloud cluster switch, so that the cloud cluster switch allocates available resources to the user according to the target demand resource information when receiving the data packet of the user, and processes the data packet based on the available resources.

[0290] For example, if the target demand resource information is 600M of bandwidth resources, the cloud cluster switch allocates 600M of bandwidth resources to the user, so that the cloud cluster switch processes the data packet based on 600M of bandwidth resources when receiving the data packet. For another example, if the target demand resource information is 15G of computing resources, the cloud cluster switch notifies the server to allocate 15G of computing resources to the user, so that the server processes the data packet based on 15G of computing resources when receiving the data packet.

[0291] In one example, for the southbound management network (network out-of-band management network), configuration and running data collection can be performed: SNMP, SysLog, Netconf, gRpc Telemetry. For collector traffic collection network, TCP collector can use ERSPAN collection, and the network side needs an IP address. The TCP collector can be deployed inside the Fabric, and the TCP collector can be deployed near the intelligent analysis module.

[0292] In one example, for the optimization and adjustment module and function, the cloud platform Manager component can respond to the analysis results of the analysis module and timely adjust the forwarding strategy of each service and issue it to each member device to fully guarantee the Qos network service for the intelligent computing center. Optimization techniques include various strategies: bandwidth guarantee, priority queue, CAR (Committed Information Rate), Qos, GTS, speed limit, etc. These are integrated and implemented on the application container of the member device.

[0293] Optimization is the allocation and optimization between various services under the condition that the entire network bandwidth meets the bandwidth demand of the intelligent computing center. If the local and core bandwidth cannot meet the requirements of the intelligent computing center, adjustment of the cloud cluster switch is required: increase the link bandwidth and increase the device expansion system processing capacity.

[0294] Using the expansion processing capacity of the cloud cluster switch, when the forwarding capacity cannot meet the demand, new member devices can be added to the cloud cluster switch. The cloud cluster switch adopts the principle of local forwarding priority, and for each member device, the local egress interface is used to forward the packet as much as possible. If the member device cannot forward the packet, other member devices will share the forwarding of the packet, thereby greatly improving the overall forwarding speed and processing capacity of the system. Each member device supports functions such as traffic caching and fast traffic switching, further improving the processing performance of the cloud cluster switch.

[0295] Using the expansion bandwidth of the cloud cluster switch, when the access devices of the cloud cluster switch (i.e. servers connected to the cloud cluster switch) increase, the service traffic increases, and the uplink bandwidth increases, new member devices can be added to increase the bandwidth of the cloud cluster switch. By adding new member devices to the cloud cluster switch, the number of cloud cluster switches does not change, and the new member devices and the existing member devices are one device, reducing the operation and maintenance problems caused by network configuration changes.

[0296] Based on the above technical scheme, the cloud cluster switch based on physical decoupling is applied to an AI intelligent data center, network operation and maintenance is placed on a cloud platform, greatly facilitating configuration management, operation and maintenance, fault troubleshooting, and dynamic expansion of equipment according to business traffic. According to the traffic and network health monitoring of the business, the device and network bandwidth and forwarding strategy are dynamically and non-perceptively optimized and adjusted. Whether the GPU is in maximum efficiency, whether it is caused by network congestion, and whether the forwarding strategy is optimized and adjusted in time due to the network.

[0297] Based on the same application concept as the above method, the embodiment of the application provides a packet processing device, which is applied to a member device in a cloud cluster switch, the member device is an NCC, an NCF or an NCP, as shown in FIG. 8A, which is a structural schematic diagram of the device, the device comprises:

[0298] A determination module 811, a receiving module 812, a processing module 813 and a sending module 814; the determination module 811 is configured to determine identity information of the member device in the cloud cluster switch;

[0299] If the identity information indicates that the member device is a first NCP, the receiving module 812 is configured to receive a first protocol packet sent by a server through a first port.

[0300] The receiving module 812 is configured to receive a first protocol packet sent by a server through a first port.

[0301] The processing module 813 is configured to add local port information of the first port to the first protocol packet to obtain a second protocol packet.

[0302] The sending module 814 is configured to send the second protocol packet to an NCC.

[0303] If the identity information indicates that the member device is an NCC, the receiving module 812 is configured to receive a second protocol packet.

[0304] The receiving module 812 is configured to receive a second protocol packet.

[0305] The processing module 813 is configured to determine global port information of the first port based on the local port information of the first port; and establish a first forwarding table item, the first forwarding table item comprising address information of the server in the first protocol packet and the global port information of the first port.

[0306] The sending module 814 is configured to send the first forwarding table item to the first NCP, so that after the first NCP receives a data packet with the address information of the server as a destination address, the first NCP queries the first forwarding table item based on the destination address to obtain the global port information of the first port, and sends the data packet through the first port.

[0307] For example, if the identity information indicates that the member device is a first NCP;

[0308] The processing module 813 is further configured to establish a first internal channel with the NCC.

[0309] The sending module 814 is further configured to send a first registration message to the NCC through the first internal channel, where the first registration message includes local port information of the first port.

[0310] For example, if the identity information indicates that the member device is a NCC;

[0311] The receiving module 812 is further configured to receive the first registration message.

[0312] The processing module 813 is further configured to configure global port information for the first port, and record a correspondence between the local port information of the first port and the global port information of the first port in a port mapping table.

[0313] For example, if the identity information indicates that the member device is a NCC;

[0314] The sending module 814 is further configured to send the first forwarding table item to a second NCP.

[0315] For example, if the identity information indicates that the member device is a second NCP;

[0316] The processing module 813 is further configured to, if a data packet with address information of the server as a destination address is received, query the first forwarding table item based on the destination address, obtain the global port information of the first port, and encapsulate the global port information of the first port for the data packet.

[0317] The sending module 814 is further configured to send the encapsulated data packet to the NCF.

[0318] For example, if the identity information indicates that the member device is a NCC;

[0319] The processing module 813 is further configured to determine global port information of a second port of a NCF, the NCF being connected to the first NCP through the second port, establish a second forwarding table item, and send the second forwarding table item to the NCF, where the second forwarding table item includes the global port information of the first port and the global port information of the second port.

[0320] If the identity information indicates that the member device is an NCF, the receiving module 812 is further configured to receive a data packet, the data packet comprising global port information of the first port; the processing module 813 is further configured to query the second forwarding table item to obtain global port information of the second port; and the sending module 814 is further configured to send the data packet through the second port.

[0321] If the identity information indicates that the member device is an NCF,

[0322] The processing module 813 is further configured to establish a second internal channel with the NCC.

[0323] The sending module 814 is further configured to send a second registration message to the NCC through the second internal channel, the second registration message comprising local port information of the second port and information of a first NCP connected to the second port.

[0324] If the identity information indicates that the member device is an NCC,

[0325] The receiving module 812 is further configured to receive the second registration message.

[0326] The processing module 813 is further configured to configure global port information for the second port and record, in a port mapping table, a correspondence between local port information of the second port, global port information of the second port and information of the first NCP.

[0327] If the identity information indicates that the member device is an NCF,

[0328] If the identity information indicates that the member device is an NCC,

[0329] The processing module 813 is further configured to establish a third internal channel with the first member device.

[0330] The receiving module 812 is further configured to receive a third registration message sent by the first member device through the third internal channel.

[0331] The sending module 814 is further configured to send the first forwarding table item to a second NCP if the first member device is the second NCP and the third registration message is received, or send the second forwarding table item to an NCF if the first member device is the NCF and the third registration message is received.

[0332] If the identity information indicates that the member device is an NCF,

[0333] if the identity information indicates that the member device is an NCC;

[0334] The processing module 813 is further configured to acquire a third forwarding table item and a fourth forwarding table item corresponding to the second member device, wherein the third forwarding table item includes address information of a server and global port information of a port on the second member device, and the fourth forwarding table item includes global port information of the port on the second member device and global port information of a port on the NCF.

[0335] The sending module 814 is further configured to send a first deletion instruction to all second NCPs, the first deletion instruction being used to make each second NCP delete the third forwarding table item corresponding to the second member device.

[0336] The sending module 814 is further configured to send a second deletion instruction to all NCFs, the second deletion instruction being used to make each NCF delete the fourth forwarding table item corresponding to the second member device.

[0337] For example, if the identity information indicates that the member device is an NCP, the NCP is the first NCP or the second NCP.

[0338] The sending module 814 is further configured to periodically send a first heartbeat message to the NCC through a first internal channel, the first heartbeat message including resource information of the NCP.

[0339] if the identity information indicates that the member device is an NCC;

[0340] The receiving module 812 is further configured to receive the first heartbeat message.

[0341] The sending module 814 is further configured to display first alarm information if the resource information of the NCP indicates that the remaining available resources of the NCP are less than a first threshold value, the first alarm information being used to prompt expansion of the NCP of the cloud cluster switch; or display second alarm information if the resource information of the NCP indicates that the remaining available resources of the NCP are greater than a second threshold value, the second alarm information being used to prompt contraction of the NCP of the cloud cluster switch.

[0342] For example, if the identity information indicates that the member device is an NCF;

[0343] The sending module 814 is further configured to periodically send a second heartbeat message to the NCC through a second internal channel, the second heartbeat message including resource information of the NCF.

[0344] if the identity information indicates that the member device is an NCC;

[0345] The receiving module 812 is further configured to receive the second heartbeat message.

[0346] The sending module 814 is further configured to display third alarm information if the resource information of the NCF indicates that the remaining available resources of the NCF are less than a third threshold value, the third alarm information being used to prompt expansion of the NCF of the cloud cluster switch; or display fourth alarm information if the resource information of the NCF indicates that the remaining available resources of the NCF are greater than a fourth threshold value, the fourth alarm information being used to prompt contraction of the NCF of the cloud cluster switch.

[0347] Based on the same application concept as the above method, an embodiment of the present application proposes a resource management device, as shown in FIG. 8B, which is a structural schematic diagram of the device. The device comprises: an acquisition module 821 configured to acquire target resource data of a cloud cluster switch, the target resource data comprising one or more of network element level resource data, network level resource data, application level resource data and service resource data; and initial demand resource information of a user, the initial demand resource information indicating expected occupied resources; a determination module 822 configured to determine target demand resource information of the user based on the target resource data and the initial demand resource information, the target demand resource information indicating actually allocated resources for the user; and a processing module 823 configured to send the target demand resource information to the cloud cluster switch, so that the cloud cluster switch allocates available resources for the user according to the target demand resource information when receiving a data packet for the user, and processes the data packet based on the available resources.

[0348] Illustratively, the network level resource data comprises resource data of the cloud cluster switch, the resource data of the cloud cluster switch comprising network overall utilization, network capacity and network health degree of the cloud cluster switch; the network element level resource data comprises resource data of each member device in the cloud cluster switch, the member device being an NCC, an NCF or an NCP; wherein the resource data of the member device comprises physical resource utilization, logical resource utilization and health degree of the member device.

[0349] Illustratively, the application level resource data comprises network quality when the cloud cluster switch forwards a data packet, the network quality comprising one or more of packet loss rate, time delay and jitter.

[0350] The obtaining module 821 obtains the application level resource data corresponding to the cloud cluster switch, and is specifically configured to: receive an RPC protocol message sent by an RPC module of a member device in the cloud cluster switch, the RPC protocol message comprising monitoring indexes counted by the member device; the member device is an NCC, an NCF or an NCP; the member device comprises an SQA module, an eMDI module and an RPC module; the SQA module sends a target traffic feature of a specified type to the eMDI module, the specified type comprising an audio type and / or a video type; if the eMDI module receives a data message matching the target traffic feature, the eMDI module counts monitoring indexes of the data message and sends the monitoring indexes to the SQA module; and the SQA module sends the monitoring indexes to the RPC module, and the RPC module generates the RPC protocol message comprising the monitoring indexes;

[0351] The application level resource data is determined based on the monitoring indexes counted by each member device.

[0352] The service resource data comprises resource data of a server accessing the cloud cluster switch; if the server is a GPU, the service resource data comprises a computing performance of the GPU and / or a memory bandwidth of the GPU;

[0353] The computing performance represents a theoretical computing capacity of the GPU when processing data of different precisions;

[0354] The memory bandwidth represents a data transmission speed provided by the GPU when processing large-scale data.

[0355] The target demand resource information is greater than the initial demand resource information, so that the available resource allocated for the user is greater than the expected occupied resource; or,

[0356] The target demand resource information is less than the initial demand resource information, so that the available resource allocated for the user is less than the expected occupied resource.

[0357] The target demand resource information is greater than the initial demand resource information, so that the available resource allocated for the user is greater than the expected occupied resource; or,

[0358] If the target resource data indicates that the available resources of the server are less than the fourth resource threshold, then the target required resource information is less than the initial required resource information, so that the available resources allocated to the user are less than the expected occupied resources.

[0359] Based on the same application concept as the above method, this application proposes an electronic device (such as a member device or management device in a cloud cluster switch, where the member device is NCC, NCF, or NCP). As shown in Figure 9, the electronic device includes a processor 911 and a machine-readable storage medium 912. The machine-readable storage medium 912 stores machine-executable instructions that can be executed by the processor 911. The processor 91 is used to execute the machine-executable instructions to implement the message processing method or resource management method of the above example of this application.

[0360] In one example, processor 911 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. Processor 911 may be implemented using at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), or PLA (Programmable Logic Array).

[0361] The processor 911 can also include a main processor and a coprocessor. The main processor, also known as the CPU (Central Processing Unit), is used to process data in the wake-up state. The coprocessor is a low-power processor used to process data in the standby state.

[0362] In some embodiments, the processor 911 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content that the display screen needs to show.

[0363] In one example, the electronic device may optionally include a peripheral device interface 913 and at least one peripheral device. The processor 911 and the peripheral device interface 913 can be connected via a bus or signal line. Each peripheral device can be connected to the peripheral device interface 913 via a bus, signal line, or circuit board. The peripheral device may include at least one of a radio frequency circuit 914 and a power supply 915.

[0364] The radio frequency circuit 914 is configured to transmit and receive RF (Radio Frequency) signals, also known as electromagnetic signals. The radio frequency circuit 914 communicates with communication networks and other communication devices using electromagnetic signals. The radio frequency circuit 914 transmits electromagnetic signals generated by electric signals and converts received electromagnetic signals into electric signals. Optionally, the radio frequency circuit 914 includes an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a user identity module card, and the like.

[0365] The radio frequency circuit 914 can communicate with user equipment through at least one wireless communication protocol. The wireless communication protocol includes, but is not limited to, the World Wide Web, a metropolitan area network, an intranet, various generations of mobile communication networks (2G, 3G, 4G, and 5G), a wireless local area network, and / or a WiFi (Wireless Fidelity) network. The power supply 915 is configured to supply power to various components in the electronic device. The power supply 915 can be an alternating current, a direct current, a disposable battery, or a rechargeable battery.

[0366] Based on the same application concept as the above method, the embodiments of the present application also provide a machine readable storage medium, and the machine readable storage medium stores a plurality of computer instructions. When the computer instructions are executed by a processor, the packet processing method or the resource management method disclosed in the above examples of the present application can be implemented.

[0367] The machine readable storage medium can be any electronic, magnetic, optical, or other physical storage device, and can contain or store information such as executable instructions, data, and the like. For example, the machine readable storage medium can be a RAM (Radom Access Memory), a volatile memory, a non-volatile memory, a flash memory, a storage drive (such as a hard disk drive), a solid state disk, any type of storage disk (such as an optical disk, a DVD, and the like), or a similar storage medium, or a combination thereof.

[0368] Based on the same application concept as the above method, the embodiments of the present application also provide a computer program product, and the computer program product can include a computer program. When the computer program is executed by a processor, the packet processing method or the resource management method is implemented.

[0369] Those skilled in the art will appreciate that embodiments of the present application can be readily used as a method, a system or a computer program product. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, embodiments of the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer-readable program code thereon for use by a computer or an instruction execution system. Program code embodied on a computer-usable storage medium can be transmitted using any

[0370] The foregoing is merely illustrative of the principles of the application and various modifications can be made by persons skilled in the art. The present application is not limited to the embodiments described above, but rather only by the claims below.

Claims

1. A method of processing a packet, the method comprising: The method is applied to a member device in a cloud cluster switch, the member device being an NCC, an NCF or an NCP, and the method comprises: determining identity information of the member device in the cloud cluster switch; if the identity information indicates that the member device is a first NCP and a first protocol packet sent by a server is received through a first port, adding local port information of the first port to the first protocol packet to obtain a second protocol packet, and sending the second protocol packet to an NCC; if the identity information indicates that the member device is the NCC and the second protocol packet is received, determining global port information of the first port based on the local port information of the first port; establishing a first forwarding table item, the first forwarding table item comprising address information of the server in the first protocol packet and the global port information of the first port; and sending the first forwarding table item to the first NCP, so that the first NCP, after receiving a data packet with a destination address being the address information of the server, queries the first forwarding table item based on the destination address to obtain the global port information of the first port, and sends the data packet through the first port.

2. The method of claim 1, wherein, Before the determination of the global port information of the first port based on the local port information of the first port, the method further comprises: if the identity information indicates that the member device is the first NCP, establishing a first internal channel with the NCC and sending a first registration message to the NCC through the first internal channel, the first registration message comprising the local port information of the first port; if the identity information indicates that the member device is the NCC and the first registration message is received, configuring global port information for the first port and recording a correspondence between the local port information of the first port and the global port information of the first port in a port mapping table.

3. The method of claim 1, wherein, After the establishment of the first forwarding table item, the method further comprises: sending the first forwarding table item to a second NCP; if the identity information indicates that the member device is the second NCP and a data packet with a destination address being the address information of the server is received, querying the first forwarding table item based on the destination address to obtain the global port information of the first port, encapsulating the global port information of the first port for the data packet, and sending the encapsulated data packet to an NCF.

4. The method of claim 3, wherein, After the determination of the global port information of the first port based on the local port information of the first port, the method further comprises: if the identity information indicates that the member device is the NCC, determining global port information of a second port of an NCF, the NCF being connected to the first NCP through the second port; establishing a second forwarding table item and sending the second forwarding table item to the NCF; wherein the second forwarding table item comprises the global port information of the first port and the global port information of the second port. If the identity information indicates that the member device is an NCF and the received data packet includes global port information of the first port, global port information of the second port is obtained by querying the second forwarding table item, and the data packet is sent through the second port.

5. The method of claim 4, wherein, Before the global port information of the second port of the NCF is determined, the method further comprises: If the identity information indicates that the member device is an NCC, a second internal channel is established with the NCC, a second registration message is sent to the NCC through the second internal channel, and the second registration message includes local port information of the second port, information of a first NCP connected to the second port. If the identity information indicates that the member device is an NCC and the second registration message is received, global port information is configured for the second port, and a correspondence relationship between the local port information of the second port, the global port information of the second port, and the information of the first NCP is recorded in a port mapping table.

6. The method of claim 1, wherein, The NCP includes an application container and a routing container, and the NCC includes an application container and a routing container; the application container of the first NCP is configured to receive the first protocol packet and send the first protocol packet to the routing container of the first NCP; the routing container of the first NCP is configured to add the local port information of the first port to the first protocol packet to obtain a second protocol packet and send the second protocol packet to the routing container of the NCC; The routing container of the NCC is configured to receive the second protocol packet and send the second protocol packet to the application container of the NCC; and the application container of the NCC is configured to determine the global port information of the first port, establish the first forwarding table item, and send the first forwarding table item to the application container of the first NCP.

7. The method of any one of claims 1-6, wherein, When the first member device joins the cloud cluster switch, the method further comprises: If the identity information indicates that the member device is an NCC, a third internal channel is established with the first member device, the third registration message sent by the first member device is received through the third internal channel; If the first member device is a second NCP and the third registration message is received, the first forwarding table item is sent to the second NCP; If the first member device is an NCF and the third registration message is received, the second forwarding table item is sent to the NCF.

8. The method of any one of claims 1-6, wherein, When the second member device leaves the cloud cluster switch, the method further comprises: ​ If the identity information indicates that the member device is an NCC and the second member device is a first NCP, a third forwarding table item and a fourth forwarding table item corresponding to the second member device are obtained; the third forwarding table item includes address information of a server and global port information of a port on the second member device, and the fourth forwarding table item includes global port information of the port on the second member device and global port information of a port on the NCF; a first deletion instruction is sent to all second NCPs, the first deletion instruction being used for causing each second NCP to delete the third forwarding table item corresponding to the second member device; a second deletion instruction is sent to all NCFs, the second deletion instruction being used for causing each NCF to delete the fourth forwarding table item corresponding to the second member device.

9. The method according to any one of claims 1 to 6, characterized in that, The method further includes: If the identity information indicates that the member device is an NCP and the NCP is the first NCP or the second NCP, a first heartbeat message is periodically sent to the NCC through a first internal channel, the first heartbeat message including resource information of the NCP; If the identity information indicates that the member device is an NCC, the first heartbeat message is received; If the resource information of the NCP indicates that a remaining available resource of the NCP is less than a first threshold value, first alarm information is displayed, the first alarm information being used for prompting expansion of the NCP of the cloud cluster switch; or if the resource information of the NCP indicates that the remaining available resource of the NCP is greater than a second threshold value, second alarm information is displayed, the second alarm information being used for prompting contraction of the NCP of the cloud cluster switch.

10. The method according to any one of claims 1 to 6, characterized in that, The method further includes: If the identity information indicates that the member device is an NCF, a second heartbeat message is periodically sent to the NCC through a second internal channel, the second heartbeat message including resource information of the NCF; If the identity information indicates that the member device is an NCC, the second heartbeat message is received; If the resource information of the NCF indicates that a remaining available resource of the NCF is less than a third threshold value, third alarm information is displayed, the third alarm information being used for prompting expansion of the NCF of the cloud cluster switch; or if the resource information of the NCF indicates that the remaining available resource of the NCF is greater than a fourth threshold value, fourth alarm information is displayed, the fourth alarm information being used for prompting contraction of the NCF of the cloud cluster switch.

11. A resource management method characterized by, The method includes: Target resource data of a cloud cluster switch is obtained, the target resource data including one or more of network element level resource data, network level resource data, application level resource data and service resource data; Initial demand resource information of a user is obtained, the initial demand resource information indicating expected occupied resources; Based on the target resource data and the initial demand resource information, target demand resource information of the user is determined, the target demand resource information indicating actually allocated resources for the user; and The method further includes: The target demand resource information is displayed. sending the target demand resource information to the cloud cluster switch, so that the cloud cluster switch allocates available resources for the user according to the target demand resource information when receiving a data packet of the user, and processes the data packet based on the available resources.

12. The method of claim 11, wherein, The network-level resource data includes resource data of the cloud cluster switch, and the resource data of the cloud cluster switch includes network overall utilization, network capacity and network health degree of the cloud cluster switch. The network element-level resource data includes resource data of each member device in the cloud cluster switch, and the member device is an NCC, an NCF or an NCP; and the resource data of the member device includes physical resource utilization, logical resource utilization and health degree of the member device.

13. The method of claim 11, wherein the application-level resource data includes network quality when the cloud cluster switch forwards a data packet, and the network quality includes one or more of a packet loss rate, a time delay and a jitter. The application-level resource data includes network quality when the cloud cluster switch forwards a data packet, and the network quality includes one or more of a packet loss rate, a time delay and a jitter. The application-level resource data includes network quality when the cloud cluster switch forwards a data packet, and the network quality includes one or more of a packet loss rate, a time delay and a jitter. The SQA module, an eMDI module and an RPC module; wherein the SQA module sends a specified type of target traffic feature to the eMDI module, and the specified type includes an audio type and / or a video type; if the eMDI module receives a data packet matching the target traffic feature, the eMDI module statistics a monitoring index of the data packet and sends the monitoring index to the SQA module; the SQA module sends the monitoring index to the RPC module, and the RPC module generates the RPC protocol packet including the monitoring index; The service resource data includes resource data of a server accessing the cloud cluster switch. If the server is a GPU, the service resource data includes a computing performance of the GPU and / or a memory bandwidth of the GPU.

14. The method of claim 11, wherein, The computing performance represents a theoretical computing capacity of the GPU when processing data of different precisions. The memory bandwidth represents a data transmission speed provided by the GPU when processing large-scale data. The target demand resource information of the user includes: If the initial demand resource information includes an expected occupied resource of the cloud cluster switch, and the target resource data indicates that available resources of the cloud cluster switch are greater than a first resource threshold, the target demand resource information is greater than the initial demand resource information, so that the available resources allocated for the user are greater than the expected occupied resource; or 15. The method according to any one of claims 11-14, characterized in that, ​ ​ If the target resource data indicates that the available resource of the cloud cluster switch is less than a second resource threshold, the target demand resource information is less than the initial demand resource information, so that the available resource allocated to the user is less than the expected occupied resource.

16. The method according to any one of claims 11-14, characterized in that, The determining the target demand resource information of the user comprises: If the initial demand resource information comprises the expected occupied resource for accessing a server of the cloud cluster switch, and the target resource data indicates that the available resource of the server is greater than a third resource threshold, the target demand resource information is greater than the initial demand resource information, so that the available resource allocated to the user is greater than the expected occupied resource; or, If the target resource data indicates that the available resource of the server is less than a fourth resource threshold, the target demand resource information is less than the initial demand resource information, so that the available resource allocated to the user is less than the expected occupied resource.

17. A packet processing device, characterized by The device is applied to a member device in a cloud cluster switch, the member device is an NCC, an NCF or an NCP, and the device comprises: a determining module, a receiving module, a processing module and a sending module; The determining module is configured to determine identity information of the member device in the cloud cluster switch. If the identity information indicates that the member device is a first NCP, The receiving module is configured to receive a first protocol packet sent by a server through a first port. The processing module is configured to add local port information of the first port to the first protocol packet to obtain a second protocol packet. The sending module is configured to send the second protocol packet to an NCC. If the identity information indicates that the member device is an NCC, The receiving module is configured to receive the second protocol packet. The processing module is configured to determine global port information of the first port based on the local port information of the first port, and establish a first forwarding table item, the first forwarding table item comprising address information of the server in the first protocol packet and the global port information of the first port. The sending module is configured to send the first forwarding table item to the first NCP, so that, after the first NCP receives a data packet with a destination address being the address information of the server, the first NCP queries the first forwarding table item based on the destination address to obtain the global port information of the first port, and sends the data packet through the first port.

18. A resource management apparatus characterized by comprising: The device comprises: An obtaining module is configured to obtain target resource data of a cloud cluster switch, the target resource data comprising one or more of network element level resource data, network level resource data, application level resource data and service resource data, and obtain initial demand resource information of a user, the initial demand resource information indicating an expected occupied resource. A determining module is configured to determine target demand resource information of the user based on the target resource data and the initial demand resource information, the target demand resource information indicating an actually allocated resource of the user. The processing module is configured to send the target demand resource information to the cloud cluster switch, so that the cloud cluster switch allocates available resources to the user according to the target demand resource information when receiving a data packet of the user, and processes the data packet based on the available resources.

19. A member device in a cloud cluster switch, the member device being an NCC, an NCF, or an NCP, characterized in that, The member device comprises a processor and a machine readable storage medium, the machine readable storage medium stores machine executable instructions capable of being executed by the processor; wherein the processor is configured to execute the machine executable instructions to implement the method of any one of claims 1-10.

20. A management device, comprising: Comprise: A processor and a machine readable storage medium, the machine readable storage medium stores machine executable instructions capable of being executed by the processor; wherein the processor is configured to execute the machine executable instructions to implement the method of any one of claims 11-16.

Images