Network measurement method for multi-modal network and in-band telemetry system
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHEJIANG LAB
- Filing Date
- 2023-04-03
- Publication Date
- 2026-06-02
Smart Images

Figure CN118784506B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of in-band network telemetry technology, specifically relating to a network measurement method and in-band telemetry system for multimodal networks. Background Technology
[0002] With the popularization and continuous development of the internet, various internet applications have emerged, posing new challenges to internet performance. On the one hand, new network architectures and design concepts have been proposed, such as the gradual transition from IPv4 to IPv6, Named Data Network (NDN), MobilityFirst, GeoNetworking, and other new network architectures. These aim to solve various problems of traditional networks, such as scalability, mobility, and security, and achieve better performance in content distribution, mobility, and security. These different network architectures constitute different network modalities. On the other hand, to adapt to the needs of different applications, technologies such as network slicing and multi-priority scheduling have become common resource scheduling and management solutions for network administrators. This also leads to the simultaneous occurrence of multiple modalities in the network. All of these factors mean that in future networks, multiple different network modalities need to coexist, making multimodal networks an important development direction for future networks.
[0003] Meanwhile, the development of programmable data planes provides network designers with a unified hardware platform and programming environment, making multimodal networks possible. Through programmable data planes, users can customize packet structure, parsing processes, and processing flows, ultimately achieving protocol-independent forwarding and laying the foundation for supporting the coexistence of multiple network modes. Furthermore, programmable data planes enable new network technologies such as in-band telemetry (INT). INT embeds measurement information into data packets, forwards them along with the packets, and the control plane ultimately collects and summarizes the measurement information. This measurement information can provide a basis for flow control in the network.
[0004] However, in networks with multiple network modes coexisting, traditional INT (Integrated Transmission) technology cannot comprehensively, accurately, and efficiently reflect the true state of the network. This is because in networks with multiple modes coexisting, different queues are needed to manage and schedule data packets of different modes in order to ensure the isolation of each mode. If traditional INT technology is used to embed measurement information in a certain mode, the measurement results cannot comprehensively and accurately reflect the information of all modes because the queues of different modes may have significant differences. If it is desired that the mode carrying measurement information records the information of all modes and all queues during measurement, it is necessary to wait for all queues to complete. Faster modes will be slowed down by slower modes, resulting in a decrease in forwarding efficiency. If measurement information is embedded in every mode, it will bring a lot of redundant information and wasted bandwidth, resulting in low efficiency. Summary of the Invention
[0005] This invention is made to solve the above-mentioned problems, and aims to provide a method for comprehensively, accurately and efficiently realizing network measurement tasks of multimodal networks, as well as an in-band telemetry system using this method.
[0006] Based on the above explanation and analysis, it can be seen that the coupling between measurement and network modes is the key factor leading to the aforementioned problems. To achieve comprehensive, accurate, and efficient network measurement in multimodal network scenarios, it is necessary to decouple measurement and network modes. Therefore, the inventors have set up a separate measurement mode for measurement tasks and assigned it the highest priority, thereby enabling efficient forwarding of measurement data packets.
[0007] Specifically, the present invention adopts the following technical solution:
[0008] This invention provides a network measurement method for multimodal networks, used for measuring multimodal networks. The method is characterized by the presence of a host and multiple programmable switches for forwarding data packets in the multimodal network. The method includes the following steps: Step S1, the host generates and sends a data packet representing the measurement mode according to a measurement instruction; Step S2, the programmable switch, acting as an intermediate node, parses the data packet upon receiving it, wherein each programmable switch has a measurement mode egress queue independent of other mode egress queues, and the measurement mode has the highest priority among all modes in the multimodal network; Step S3, the programmable switch determines whether the data packet belongs to the measurement mode based on the parsing result; Step S4, if the determination is yes in Step S3, the programmable switch extracts the corresponding target measurement information from its storage space according to the measurement instruction and embeds it into the data packet; Step S5, the programmable switch places the data packet with embedded information into the measurement mode egress queue and forwards it.
[0009] The network measurement method for multimodal networks provided by the present invention may also have the following technical features: in step S2, the programmable switch extracts the mode identification field of the received data packet; in step S3, the programmable switch determines whether the data packet belongs to the measurement mode based on the content of the extracted mode identification field.
[0010] The network measurement method for multimodal networks provided by the present invention may also have the following technical features, wherein the data packet of the measurement mode has a measurement header field, and in step S4, the programmable switch embeds the target measurement information after the measurement header field of the data packet.
[0011] The network measurement method for multimodal networks provided by the present invention may also have the following technical features, including the following steps: Step S6, when the determination in step S3 is negative, the programmable switch puts the data packet into the egress queue corresponding to its mode, forwards it, and records the information of the data packet in the storage space of the programmable switch.
[0012] The network measurement method for multimodal networks provided by the present invention may also have the following technical feature: the measurement mode exit queue is connected to the next-hop programmable switch, router or host.
[0013] The network measurement method for multimodal networks provided by the present invention may also have the following technical features, wherein the target measurement information includes switch ID, data packet arrival timestamp, queue length and queuing delay for each of the modes.
[0014] This invention provides an in-band telemetry system, set up in a multimodal network, for measuring the multimodal network. The system comprises: a host; and multiple programmable switches for forwarding data packets. Each programmable switch has a measurement mode egress queue independent of other mode egress queues, and the measurement mode has the highest priority among all modes in the multimodal network. According to a measurement instruction, the host generates and sends a measurement mode data packet. When a programmable switch, acting as an intermediate node, receives a data packet, it parses the packet and determines whether the packet belongs to the measurement mode based on the parsing result. If it does, the programmable switch extracts the corresponding target measurement information from its storage space according to the measurement instruction and embeds it into the data packet. The embedded data packet is then placed into the measurement mode egress queue for forwarding.
[0015] Invention Function and Effect
[0016] According to the network measurement method and in-band telemetry system for multimodal networks of the present invention, programmable switches are used as measurement nodes. Each programmable switch has a measurement mode exit queue independent of other mode queues, and the measurement mode has the highest forwarding priority. For data packets of the measurement mode, the target information stored in the programmable switch is embedded. For ordinary data packets of other modes, they are forwarded normally and their data is recorded. That is, the method of the present invention decouples the measurement function from other network modes in the multimodal network. The solution is to treat measurement as a separate mode. Therefore, measurement data does not need to wait in the queues of various mode data packets, but can be forwarded directly in the fast channel. At the same time, it can carry all the information of the queues corresponding to all other network modes, thereby realizing comprehensive, accurate and efficient network measurement in multimodal scenarios. At the same time, isolating the measurement mode and using it only for performing measurement tasks also effectively avoids the problem of excessively large data packet size and exceeding the maximum transmission unit that may be caused by in-band telemetry technology, allowing measurement mode data packets to carry more measurement information. Attached Figure Description
[0017] Figure 1 This is a flowchart of a network measurement method for multimodal networks in an embodiment of the present invention;
[0018] Figure 2 This is a schematic diagram of the data packet forwarding process in an embodiment of the present invention;
[0019] Figure 3 This is a schematic diagram of a multimodal network in an embodiment of the present invention, showing the coexistence of various network traffic types. Detailed Implementation
[0020] To make the technical means, creative features, objectives and effects of the present invention easy to understand, the network measurement method and system for multimodal networks of the present invention will be specifically described below with reference to embodiments and accompanying drawings.
[0021] <Example>
[0022] This embodiment provides a network measurement method for multimodal networks, used to measure multimodal networks. As one example, it is used to measure multimodal networks where multiple network traffic types coexist.
[0023] Figure 3 This is a schematic diagram of a multimodal network in this embodiment where multiple network traffic types coexist.
[0024] like Figure 3As shown, in this multimodal network, data packets from different network architectures coexist within the same network, including IPv4, IPv6, NDN, GeoNetworking, and MobilityFirst packets, each corresponding to a network mode. These network modes have different application layer requirements. For example, NDN, due to its close connection with content delivery, typically has long-running data streams with high bandwidth requirements but lower real-time requirements, resulting in relatively longer queuing latency. GeoNetworking and MobilityFirst, on the other hand, have higher mobility requirements and therefore higher real-time requirements, resulting in relatively shorter queuing latency.
[0025] In this embodiment, the multiple switches in the multimodal network are all programmable switches, which also serve as measurement nodes. Each programmable switch is equipped with a computer program for implementing the measurement method of this embodiment. A measurement mode, independent of the other modes in the multimodal network, is provided; that is, the measurement mode has a separate egress queue in each programmable switch. Among all modes in the multimodal network, the measurement mode has the highest priority, meaning that data packets from the measurement mode will be forwarded with the highest priority.
[0026] Each programmable switch's storage space will record information about every data packet arriving at the switch, including the packet's arrival timestamp, etc. This is a standard setting and will not be elaborated upon further.
[0027] In this embodiment, the programmable switch is model Barefoot Tofino.
[0028] The data packet has a mode identifier field, which determines which mode the data packet belongs to. Compared with ordinary data packets (i.e., data packets of other modes), the header format of measurement mode data packets further includes a measurement header field, after which the target measurement information (the content information to be measured) is embedded.
[0029] Figure 1 This is a flowchart of the network measurement method for multimodal networks in this embodiment.
[0030] like Figure 1 As shown, the network measurement method for multimodal networks specifically includes the following steps:
[0031] Step S1: According to the measurement command, the host generates and sends out a data packet of the measurement mode.
[0032] In this embodiment, the network administrator configures the measurement task on the telemetry server (host), including configuring the target measurement content, measurement frequency, etc.; then, the telemetry server sends the corresponding measurement instructions to each programmable switch and sends data packets for the measurement mode. The measurement content may include, but is not limited to: the switch ID through which the data flow passes, the data packet arrival timestamp, the queue length corresponding to each network mode (i.e., the length of the switch port queue), and the queuing delay, etc.
[0033] In other words, the purpose of the measurement task is the same as that of traditional network measurement: to detect the size of data flow in the network, identify network congestion, and verify the consistency between the data plane and the control plane, thereby providing a basis for congestion control and flow engineering solutions. Therefore, the measurement task itself and measurement instructions will not be elaborated upon further.
[0034] For ease of understanding, in this embodiment, it is assumed that there are n modes, denoted as mode 1, mode 2, ..., mode n, and the main measurements are the queue length and queuing delay of these n modes.
[0035] Step S2: When the programmable switch, acting as an intermediate node, receives a data packet, it parses the data packet.
[0036] Step S3: The programmable switch determines whether the received data packet belongs to the measurement mode based on the parsing result.
[0037] Specifically, the programmable switch, acting as an intermediate node, extracts the mode identifier field of the received data packet, determines the mode to which the data packet belongs based on the content of the field, and performs different processing on measurement mode data packets belonging to the measurement mode and ordinary data packets of other modes.
[0038] Step S4: When the determination in step S3 is yes, the programmable switch extracts the corresponding target measurement information from its storage space according to the measurement instruction and embeds the target measurement information into the measurement mode data packet.
[0039] That is, according to the content to be measured as indicated by the measurement command, the programmable switch selects the corresponding partial information from its storage space and embeds it into the measurement header field of the measurement mode data packet.
[0040] In this embodiment, based on the above assumptions, after the programmable switch extracts information such as queue length and queuing delay of n modes from its storage space and embeds it into the measurement header field of the measurement mode data packet, the embedded information includes:
[0041] Switch ID (32-bit)
[0042] Data packet arrival timestamp (32 bits)
[0043] Mode 1 queue length (32 bits)
[0044] Mode 1 queuing delay (32-bit)
[0045] Modal 2 queue length (32 bits)
[0046] Modal 2 queuing delay (32-bit)
[0047] ...
[0048] Length of modal n queue (32 bits)
[0049] Mode n queuing delay (32-bit)
[0050] In step S5, the programmable switch places the measurement mode data packet containing the target measurement information into the egress queue corresponding to the measurement mode and forwards it.
[0051] Figure 2 This is a schematic diagram of the forwarding process on the programmable switch in this embodiment. The blank squares represent measurement mode data packets, while the two striped squares represent data packets of two other network modes, respectively.
[0052] like Figure 2 As shown and as mentioned above, the measurement mode has a separate egress queue, and because it has the highest priority, the measurement mode data packets in this egress queue will be forwarded first, which can prevent the measurement data packets from entering the queues corresponding to other modes to wait, thereby ensuring efficient measurement.
[0053] In addition, the outbound queue connects to a next-hop switch, router, or host. When connected to a next-hop switch, the next-hop switch acts as a measurement node and repeats steps S2 to S5.
[0054] After the measurement mode data packet arrives at its destination, the destination host will parse it and forward the target measurement information it carries to the control plane. The control plane will then summarize and process the measurement information. In other words, the measurement task itself, as well as the further processing of the measurement mode data packet after it arrives at its destination, are consistent with existing network measurement methods, and therefore will not be elaborated further.
[0055] Step S6: If the determination in step S3 is negative, that is, if the received data packet is a normal data packet of other modes, the programmable switch puts the normal data packet into the egress queue corresponding to its mode, forwards it, and records the information of the normal data packet in its storage space.
[0056] That is, the programmable switch stores information about the data packets arriving at the switch and updates this information in real time. In step S4 above, the specified measurement information is extracted from this stored information.
[0057] This embodiment also provides an in-band telemetry system corresponding to the above method. The system includes multiple programmable switches as described above. In this system, similarly, the programmable switch acting as the source node sends a data packet of measurement mode according to the measurement command. The programmable switch acting as the intermediate node parses the received data packet, determines whether it is a data packet of measurement mode, and performs different processing on the data packet of measurement mode and ordinary data packet, as described above, and will not be repeated.
[0058] In this embodiment, the parts not described in detail are well-known technologies in the art.
[0059] Functions and effects of the embodiments
[0060] According to the network measurement method and in-band telemetry system for multimodal networks provided in this embodiment, programmable switches are used as measurement nodes. Each programmable switch has a measurement mode exit queue independent of other mode queues, and the measurement mode has the highest forwarding priority. For data packets of the measurement mode, the target information stored in the programmable switch is embedded. For ordinary data packets of other modes, they are forwarded normally and their data is recorded. That is, the method of this embodiment decouples the measurement function from other network modes in a multimodal network where multiple network traffics coexist. The solution is to treat measurement as a separate mode. Therefore, measurement data does not need to wait in the queues of various mode data packets, but can be forwarded directly in the fast channel. At the same time, it can carry all the information of all other network mode corresponding queues, thereby realizing comprehensive, accurate, and efficient network measurement in multimodal scenarios. Meanwhile, isolating the measurement mode and using it only for performing measurement tasks also effectively avoids the problem of excessively large data packet size and exceeding the maximum transmission unit that may be caused by in-band telemetry technology, allowing measurement mode data packets to carry more measurement information.
[0061] The above embodiments are only used to illustrate specific implementations of the present invention, and the present invention is not limited to the scope of the description of the above embodiments.
[0062] In the above embodiments, for ease of understanding, a multimodal network with multiple network traffic coexisting was used as an example for illustration. However, the method of the present invention is not limited to this. The present invention does not limit which modes are specifically included in the multimodal network, and the method of the present invention can also be applied to multimodal networks with other modes.
[0063] Furthermore, in order to adapt to the needs of different applications, technologies such as network slicing and multi-priority scheduling have emerged. Different network slices and different priorities are logically equivalent to multiple network modes in modeling because they have different priorities, speeds, bandwidths, and real-time requirements. The method of this invention can also be applied to these technologies after corresponding modifications.
[0064] In the above embodiments, for ease of understanding, the measurement information includes switch ID, packet arrival timestamp, queue length and queuing delay for each network mode. However, the method of the present invention does not limit the measurement information. Depending on the actual needs of the network administrator, the measurement information may include only some of the above information or other target content information.
[0065] In the above embodiment, the programmable switch model is Barefoot Tofino. In alternative solutions, other models of programmable switches may also be used.
Claims
1. A network measurement method for multimodal networks, used for measuring multimodal networks, characterized in that, The multimodal network has a host and multiple programmable switches for forwarding data packets. The method includes the following steps: Step S1: According to the measurement instruction, the host generates and sends a data packet of the measurement mode, which includes a measurement header field; Step S2: When the programmable switch, acting as an intermediate node, receives a data packet, it parses the data packet. Each programmable switch has a measurement mode egress queue that is independent of other mode egress queues. The measurement mode is independent of other network modes in the multimodal network, and the measurement mode has the highest priority among all modes in the multimodal network. Step S3: The programmable switch determines whether the data packet belongs to the measurement mode based on the parsing result; Step S4: When the determination in step S3 is yes, the programmable switch extracts the corresponding target measurement information from its storage space according to the measurement instruction and embeds the target measurement information after the measurement header field of the data packet. The target measurement information includes: the switch ID through which the data flow passes, the data packet arrival timestamp, the queue length and queuing delay corresponding to each network mode. In step S5, the programmable switch places the data packet containing the target measurement information into the measurement mode egress queue and forwards it.
2. The network measurement method for multimodal networks according to claim 1, characterized in that: in, In step S2, the programmable switch extracts the mode identifier field of the received data packet. In step S3, the programmable switch determines whether the data packet belongs to the measurement mode based on the content of the extracted mode identifier field.
3. The network measurement method for multimodal networks according to claim 1, characterized in that, It also includes the following steps: Step S6: If the determination in step S3 is negative, the programmable switch places the data packet into the egress queue corresponding to its mode, forwards it, and records the information of the data packet in the storage space of the programmable switch.
4. The network measurement method for multimodal networks according to claim 1, characterized in that: in, The measurement mode exit queue is connected to the next-hop programmable switch, router, or host.
5. An in-band telemetry system, configured in a multimodal network, for measuring the multimodal network, characterized in that, include: Host; as well as Multiple programmable switches are used to forward data packets. Each of the programmable switches has a measurement mode egress queue that is independent of other mode egress queues. This measurement mode is independent of other network modes in the multimodal network, and among all network modes in the multimodal network, the measurement mode has the highest priority. According to the measurement instructions, the host generates and sends a data packet in the measurement mode, which includes a measurement header field. When the programmable switch, acting as an intermediate node, receives a data packet, it parses the packet and determines whether the packet belongs to the measurement mode based on the parsing result. If it does, the programmable switch extracts the corresponding target measurement information from its storage space according to the measurement instruction, embeds the target measurement information into the measurement header field of the data packet, and places the data packet with the embedded target measurement information into the measurement mode egress queue for forwarding. The target measurement information includes: the switch ID through which the data flow passes, the data packet arrival timestamp, the queue length and queuing delay corresponding to each network mode.