142results about How to "Reduce forwarding delay" patented technology

The invention discloses a data center network routing method based on a Fat-Tree structure, and belongs to the technical field of data center networks. The SDN technology is used by the method, the data stream size and the link transmission condition are comprehensively considered in a data center network of a Fat-Tree structure and having multipath transmission ability, and new streams entering the network are routed by using a dynamic load balancing (DLB) algorithm based on residual link bandwidth and the number of large streams; and after link congestion is detected, the large streams on the congested link are re-scheduled by using a dynamic re-routing (DR) algorithm based on deadline and switch queue length from large to small, until the link is not congested. By adoption of the data center network routing method disclosed by the invention, the load balance of data streams in the data center network can be realized, the routing strategy can be dynamically adjusted according to the network state, the utilization rate and the throughput of the link are improved, and the forwarding delay and the packet loss rate of the data streams are reduced.

The invention provides a method and device for detecting the message sending rate. The method comprises the steps that the cache size of an interchanger is monitored; whether the catch state of the interchanger is about to reach the maximum extent or not is judged according to the cache size; when the catch state of the interchanger is about to reach the maximum extent, whether the priority of a message to be forwarded is high or low is judged; the interchanger receives the message or directly discards the message according to the priority judgment on the message. According to the method, the cache size of the interchanger is monitored in real time, when the catch size reaches the preset cache threshold value of a system, the interchanger is marked, the message with high priority can occupy the cache of the interchanger at any time, a message with low priority occupies the cache space of the interchanger only when the cache of the interchanger is not used up, and therefore the forwarding time delay of the messages can be shortened in the forwarding process.

The invention discloses a virtual connection supporting real-time embedded gateway based on a controller area network (CAN) bus and Ethernet. The real-time embedded gateway comprises hardware, a supporting module and a virtual connection module. A plurality of embedded development boards simultaneously with CAN bus interfaces and Ethernet interfaces are adopted, and are connected by the CAN bus to form a virtual connection service supporting trunking system. The supporting module of the gateway consists of a driver, a mu c-OS II operating system and a light weight Internet protocol (LWIP) protocol stack. The virtual connection module comprises a virtual connection protocol sub-module and a virtual connection state sensing sub-module, and is used for finishing actual virtual connection forwarding and maintenance work. The real-time embedded gateway has the advantages of configuration flexibility, high adaptability, high extensibility, low energy consumption and the like, and is applied to data acquisition and transmission systems of relatively larger scales.

In the stack chain constituted by exchange chips, the normal forwarding flow of the data frames received by exchange chips passing through is as follows. (A) Data frames are received. (B) The protocol process of the frame and looking up the forwarding target port of the frame. (C) Determining whether the target port is the local port. If yes, the frame is sent to the target port through the flow (D). The characters are that before the flow (B), the stack process branch is setup to determine whether the source port is the stack port. At least for the data frame, whose source port is the stack port, the determination of whether the current chip is the end of the stack chain is carried out. If the current chip is not at the end of the stack chain, the data is sent to next stack port quickly.

The present application provides a method and an apparatus for configuring and delivering a flow table entry. The method comprises: a switch sending, to a controller, a request message for requesting delivering a flow table entry, the request message carrying a to-be-forwarded data packet; after the controller receives the request message, the controller generating all flow table entries matching the to-be-forwarded data packet, and returning, to the switch, a response message carrying all the flow table entries matching the to-be-forwarded data packet; after then switch receives the response message, the switch extracting all the flow table entries matching the to-be-forwarded data packet from the response message; and aggregating all the extracted flow table entries into flow table entries of a hardware layer, and when the switch supports the flow table entries of the hardware layer, the switch configuring the flow table entries of the hardware layer into a flow table of the hardware layer. The method and the apparatus for configuring and delivering a flow table entry in the present application can shorten the time of configuring flow table entries of a hardware layer into a flow table of the hardware layer, can shorten the delay of forwarding the data packets, and can reduce the number of lost packets of the data packets.

The application discloses a method and a device for processing messages, in order to solve the problem that ultra-low-time-delay forwarding requirement can not be met due to large forwarding time delay of time-delay-sensitive business. The method comprises the steps of: acquiring a first slicing data frame included in a first data frame from an MAC (Media Access Control) layer by first network equipment, caching the first slicing data frame to a first queue; acquiring a second slicing data frame included in a second data frame from the MAC layer by the first network equipment, caching the second slicing data frame to a second queue; transmitting the first slicing data frame to a forwarding processing module by the first network equipment, and acquiring first forwarding information throughthe forwarding processing module; and transmitting the second slicing data frame to the forwarding processing module by the first network equipment after the first slicing data frame is transmitted tothe forwarding processing module, and acquiring second forwarding information through the forwarding processing module. Through adoption of the method provided by the application, forwarding time delay of the time-delay-sensitive business in network equipment can be reduced and the time-delay-sensitive business can not be affected by non-time-delay-sensitive business.

The provided message transmission method comprises: when there is congestion on link, receiving LLQ message, deciding whether the LLQ is idle and meets condition for adding sequence; if yes, continual deciding whether the LLQ message meets the condition for out of sequence, and directly transmitting the message if also yes. This invention can reduces transmission time delay.

The invention discloses a time-triggered Ethernet switch based on a biplane and a packet switching method, which mainly solve the problem that the deterministic and real-time performances of the existing time-triggered Ethernet switch forwarding a time-triggered TT service are poor. The device comprises an input shunting module (1), a clock synchronizer (2), a time-triggered TT switching plane (3), an event-triggered ET switching plane (4) and an output arbitration module (5). The input shunting module is respectively connected with the clock synchronizer, the TT switching plane, and the ET switching plane, to complete switch clock synchronization and to forward the TT service frame and the ET service frame respectively; the TT switching plane and the ET switching plane are both connectedwith the output arbitration module to select a service frame which is preferentially sent and to complete sending of data. The time-triggered Ethernet switch based on a biplane and the packet switching method in the invention can ensure absolute priority forwarding of the time-triggered TT service, reduce the forwarding delay of the time-triggered TT service, and can be applied to a time-triggeredswitching network with high speed and low delay.

The invention provides an issuing method and device of a multicast forwarding table item. The issuing method comprises the steps of receiving a multicast broadcast, obtaining the corresponding multicast forwarding table item, and judging whether the multicast forwarding table item records an outbound interface or not; judging whether the multicast broadcast is a trans-board forwarding broadcast or not according to a preset rule if there is a record of the outbound interface in the multicast forwarding table item; issuing the multicast forwarding table item to an interface board where the outbound interface exists, an interface board where an inbound interface exists, and a forwarding chip of a main control panel if the multicast broadcast is a trans-board forwarding broadcast; issuing the multicast forwarding table item to a forwarding chip of the interface board where the outbound interface exists if the multicast broadcast is not a trans-board forwarding broadcast. According to the conditions of trans-board forwarding and non-trans-board forwarding broadcasting, the interface board or the forwarding chip of the main control panel irrelevant to the multicast forwarding table item are not used for issuing the multicast forwarding table item, thus waste of the forwarding chip resources are lowered, and forwarding time delay of a frame-type device is also shortened.

The invention discloses a large capacity and expandable grouping exchange network structure, which adopts three level structure, wherein a first level is a flow distribution level, a second level is a transfer level consisting of a plurality of separated loop network, and a third level is a transaction flow convergence level. The structure can fully utilize the middle level loop network to accomplish multicast, and provides a good network property and expansion capacity, so as to be suitable for serving as an exchange network for the grouping internet core router.

The embodiment of the invention discloses a message matching system and method based on NFV. NFV network element equipment sends packaged rule data to a hardware accelerator. Wherein the rule data isa parameter on which message matching depends. The hardware accelerator receives the packaged rule data and stores the analyzed rule data to a preset position. When the hardware accelerator receives anetwork message, data information carried by the network message is analyzed; and the hardware accelerator receives the data information sent by the NFV network element device and queries target ruledata matched with the data information from a preset position, and the hardware accelerator can feed back the matched target rule data to the NFV network element device, so that the NFV network element device processes the network message according to the target rule data. The external hardware accelerator is introduced into an NFV system to realize an NAT function, so that the CPU load of NFV network element equipment is reduced, the message forwarding efficiency is improved, and the message forwarding time delay is shortened.

The invention discloses a multi-route inter-satellite communication link based on cluster formation. The multi-route inter-satellite communication link based on the cluster formation is composed of acommunication antenna which is exclusively used for omnidirectional signal transmitting and receiving; inter-satellite terminal machines which are exclusively used for multichannel searching/capturing/capturing; and processor which are exclusively used for inter-satellite data transmitting and receiving and screening. According to the multi-route inter-satellite communication link, through adoption of a multichannel parallel capturing technology, the same cluster satellite signal is searched; through utilization of a queue type multi-route communication technology, inter-satellite informationof all satellites in a cluster is directly or indirectly switched; through adoption of a forwarding delay dynamic optimization technology, the processors determine transmitting opportunity according to priority of current data and reduce forwarding delay of important data to the lowest; and a multiway information optimization technology is employed. According to the multi-route inter-satellite communication link, the problem that a traditional inter-satellite communication scheme is few in access terminal number, serious in dependence on antenna visibility and long in forwarding delay is solved.

The invention discloses an SDN (Software Defined Network) intelligent routing data transmission method based on distributed deep reinforcement learning, which realizes calculation of a fast routing path, maximizes throughput under the condition of ensuring delay, and solves the problems of low speed and small throughput of the traditional algorithm. According to the method, a reinforcement learning algorithm is used, the routing calculation process is simplified into simple input and output through the algorithm, multiple iterations during calculation are avoided, so that rapid calculation ofthe routing path is achieved, forwarding delay is reduced through acceleration of the speed of the routing algorithm, and data packets originally abandoned due to ttl expiration have higher probability of survival and are forwarded successfully so that network throughput is increased. During two training stages of offline training and online training is set, parameters are updated in a dynamic environment to select an optimal path, so that the invention has topology adaptability.

The invention provides a method for reducing NVR live forwarding time delay. The method is applied to an NVR. The method includes the first step of receiving a live video-on-demand request from a monitoring rear-end device and judging whether the on-demand live flow is recorded in a live flow information sheet, if not, receiving the IP address of the interface receiving the live video-on-demand request and the IP address of the monitoring rear-end device, sending the IP address of the interface of the live video-on-demand request and the IP address of the monitoring rear-end device to an on-demand monitoring front-end device and establishing a corresponding live flow information sheet, and the second step of carrying out forwarding processing through an exchange chip of a local network card according to the target IP address of a message after receiving the on-demand live flow sent by the monitoring front-end device, wherein the target IP address of the live flow sent by the monitoring front-end device is the IP address of the monitoring rear-end device, and a source IP address is the IP address of the interface receiving the live video-on-demand request. Through the method, the NVR live forwarding time delay is effectively reduced, and performance pressure of the NVR is relieved.

A dynamic bandwidth allocation method of network slices and an OLT relate to a DBA method of a PON system, and the method comprises the steps of a1, setting a service quality level of each network slice, and establishing a corresponding relationship with T-CONT; configuring the remaining bandwidth scheduling type of the T-CONT to be none, NA or BE; a2, according to the bandwidth requirement collected in the last DBA period of each T-CONT, allocating the bandwidth of the current DBA period to the network slice, wherein the bandwidth allocated to each T-CONT is not greater than the sum of the fixed bandwidth and the guaranteed bandwidth; a3, if the PON port has the unallocated bandwidth, sequentially meeting the bandwidth requirements of the T-CONT with the remaining bandwidth scheduling type of NA according to the service quality level; and a4, if the PON port still has the unallocated bandwidth, sequentially meeting the bandwidth requirements of the T-CONT with the remaining bandwidthscheduling type of BE according to the service quality level. According to the invention, differentiated quality guarantee services are provided for different network slice users.

The invention discloses an SDN (Software-Defined Network) forwarding system and method. The system comprises a controller, and at least two switches which are arranged between a first host requiring communication and a second host requiring to be communicated, wherein the controller is used for determining forwarding path information of a target data packet and transmitting the forwarding path information to the first switch connected with the first host; after the first host receives the target data packet, a forwarding path label transmitted by the controller is added for the target data packet; each switch is used for determining a corresponding forwarding physical address and a corresponding forwarding port according to the forwarding path label and forwarding the target data packet according to the corresponding forwarding physical address and the corresponding forwarding port, so that the forwarding of the target data packet from the first host to the second host is finished; and the target data packet comprises information of physical addresses and forwarding ports of all the switches in a forwarding path, and data packet forwarding can be carried out directly according to the information after each switch receives the target data packet, so that forwarding time delay is reduced and the forwarding efficiency is increased.

The invention discloses a dynamic virtual private network (DVPN) based data transmission method and a device. The method includes that a Spoke sends a Cost Request message to a neighbor Spoke; the Spoke receives a Cost Response message returned by the neighbor Spoke, and a Cost value between the neighbor Spoke and a target Spoke is carried in the Cost Response message; and the Spoke compares a Cost value between the device and the target Spoke with Cost values among the device, the neighbor Spoke and the target Spoke, and when data are required to be sent to the target Spoke, the data are sent to the target Spoke through a path with the minimum Cost value. According to the DVPN based data transmission method and the device, the forward delay of the data can be reduced.

The invention discloses a low-delay micro-service routing management system, which reduces message forwarding delay, supports cross-LAN deployment and meets fault tolerance requirements. According tothe technical scheme, service discovery, service offline and service routing functions in a micro-service cluster are supported. Distributed management is used in the service routing table, and the routing table of each node can be automatically adjusted along with service and node changes in the cluster. Single-LAN and multi-LAN deployment is supported, and all nodes in the same domain exchange routing table records. A plurality of LANs are cascaded by deploying proxy nodes, the proxy nodes exchange routing information with one another through a TCP network and inform nodes in the domain, andeach node in the cluster can obtain the routing information of the whole cluster. Distributed routing management is used, after a server or a proxy node exits accidentally, expected connected nodes can automatically perform serve offline processing and inform other nodes in the cluster, and the cluster automatically completes routing adjustment under node faults.

The embodiment of the invention discloses a stack splitting processing method and device, which are used for reducing the service interruption time in the stack splitting process, improving the flow switching performance and improving the user experience. First, a first stacking member device acquires a working state of an uplink service port of each stacking member device in a stacking system, the uplink service port refers to a service port of each stacking member device in the stacking system connected with an upstream device, and the stacking system comprises the first stacking member device and a second stacking member device; secondly, the first stacking member device determines a target device according to the working state of the uplink service port of each stacking member device in the stacking system; and finally, when the first stacked member device is determined as the target device, the first stacked member device executes a port closing operation, and the port closing operation comprises switching all service ports in a working state of the first stacked member device to a non-working state.

The embodiment of the invention discloses a method and a device for realizing load balancing. A specific embodiment of the method comprises the following steps: acquiring a server information list; responding to the received access request, and determining a target area from the at least one area; based on the target area, executing the following polling steps: polling a plurality of nodes in thetarget area according to a balance tree corresponding to the target area, and determining whether the target area comprises available nodes or not; in response to determining that the target area comprises available nodes, taking the obtained first available node as a target node; sending the access request to a server corresponding to the target node; and in response to determining that the target area does not include the available node, determining a new target area according to the server information list, and continuing to execute the polling step. According to the embodiment, the forwarding delay is reduced, and the forwarding efficiency is improved.

The invention provides an encounter probability-based adaptive opportunistic forwarding method applied in a vehicular Ad Hoc network. The method comprises the steps that any one vehicle is selected from the vehicular network to serve as a request vehicle node, and the request vehicle node is set to generate a message and need to forward the message to a destination node; if an encounter probability prediction value of a candidate forwarding node and the destination node is greater than that of the request vehicle node and the destination node, the step 3 is executed; the encounter lasting timeof the request vehicle node and the candidate forwarding node is predicted, and if the encounter lasting time of the request vehicle node and the candidate forwarding node is longer than a thresholdTmin, the request vehicle node forwards the message to the candidate forwarding node; if the message forwarding time is longer than a preset updating cycle, the message is forwarded to a roadside unitand then forwarded to a vehicle node passing through the roadside unit by the roadside unit; and if the message forwarding time T is longer than an updating cycle, the request vehicle node continuesto move.