Light path inspection system and method

Abstract

The invention relates to the technical field of optical fiber communication, and provides an optical path inspection system and method.The optical path inspection system comprises an optical switch, a leaf switch, an optical fiber link, a controller and an acquisition system platform, the optical fiber link is provided with a server, and the optical switch is provided with a total acquisition port, a forwarding port and an output port; the total acquisition ports, the forwarding ports and the optical fiber links are the same in number, the total acquisition ports are connected with the optical fiber links, the forwarding ports are connected with the leaf switches, and the output ports are connected with the acquisition system platform; the controller is responsible for converting the inspection requirement of the acquisition system platform into a configuration rule of the light exchanger and issuing the configuration rule, and monitoring the state of the light exchanger to inform the acquisition system platform when to start acquisition. According to the invention, the collection cost is effectively reduced, the number of collection ports does not need to be configured according to the number of optical fiber links in a ratio of 1: 1, and all information is ensured to be collected in the inspection collection period, so that high collection resolution can be provided.

Description

Technical Field

[0001] The present application relates to the field of optical fiber communication technology, and in particular to an optical path inspection system and method. Background Art

[0002] As a soft collection method, the traditional collection solution based on Telemetry technology has many defects. Telemetry technology usually occupies the performance resources of the switch, which in turn affects business efficiency. At the same time, due to the limitation of the hardware performance of business equipment, the monitoring time granularity of this type of solution is relatively coarse, usually only reaching the millisecond level, with low resolution, and it is difficult to capture more subtle events in the network. In addition, Telemetry technology can usually only collect statistics or status information of the link, and the information provided is limited. Summary of the invention

[0003] In order to help solve the above technical problems, the present application provides an optical path inspection system and method.

[0004] In a first aspect, the present application provides an optical path inspection system, which adopts the following technical solution: An optical path inspection system, wherein the optical path inspection system includes an optical switch, a leaf switch, at least one optical fiber link, a controller, and a collection system platform. A server is provided on the optical fiber link, and the server is used to transmit the collected information to the leaf switch; The optical switch is provided with at least one full-quantity collection port, at least one forwarding port and an output port. The number of full-quantity collection ports, forwarding ports and optical fiber links is the same. The full-quantity collection port is connected to the optical fiber link for full-quantity collection of collection information. The forwarding port is connected to the leaf switch for forwarding the collection information to the leaf switch. The output port is connected to the collection system platform for outputting the inspected collection information to the collection system platform.

[0005] The optical path inspection system also includes a controller. There is a bidirectional port connection between the controller and the acquisition system platform, and there is a bidirectional port connection between the controller and the optical switch. The acquisition system platform sends an inspection plan to the controller. The controller is used to output configuration rules according to the inspection plan. The optical switch executes the configuration rules. At the same time, the optical switch returns the status information after executing the configuration rules to the controller. The controller decides when to notify the acquisition system platform to start collecting signals based on the status information.

[0006] Preferably, the inspection scheme includes inspection frequency, inspection duration and inspection object, and the configuration rule includes switching the inspection link.

[0007] Preferably, the optical path inspection system also includes a spine switch, which is connected to the output end of the leaf switch and is used to receive the collected information.

[0008] Preferably, the server is a computing device equipped with a computing unit, and the computing unit includes a GPU.

[0009] In a second aspect, the present application provides a light path inspection method of the light path inspection system as described in the first aspect, which adopts the following technical solution: An optical path inspection system, wherein the optical path inspection method comprises: The full data collection port collects the data of all optical fiber links; The forwarding port forwards the collection information of all optical fiber links to the leaf switch, and the output port outputs the collection information of one optical fiber link to the collection system platform; The acquisition system platform sends the inspection plan to the controller, and the controller outputs the configuration rules to the optical switch according to the inspection plan; The optical switch executes the configuration rules and returns status information to the controller; After receiving the status information of the optical switch completing the switching, the controller notifies the collection system platform that it can start collecting information.

[0010] Preferably, the optical path inspection method further includes: The collection system platform collects and receives the collection information, determines the failed optical fiber link according to the collection information, and controls the optical switch to fully collect the collection information of the failed optical fiber link.

[0011] Preferably, the inspection scheme includes inspection frequency, inspection duration and inspection object, and the configuration rule includes switching the inspection link.

[0012] Preferably, the optical switch executes a configuration rule including: The output port outputs the collection information of another optical fiber link to the collection system platform.

[0013] Preferably, the acquisition system platform sends an inspection plan to the controller, and the controller outputs configuration rules to the optical switch according to the inspection plan, including: The acquisition system platform can send the inspection plan to the controller in the form of a predefined rule interface through the control plane network. The controller translates the inspection plan into configuration rules and sends the configuration rules to the optical switch through the management interface.

[0014] In summary, the optical path inspection system of the present application effectively reduces the cost of collection and ensures full collection of all information during the inspection and collection period, thereby providing high collection resolution and more comprehensive link information. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic diagram of the network topology architecture of the relevant technology intelligent computing center; Figure 2A schematic block diagram of an embodiment of an optical path inspection system of the present application; Figure 3 This is a partial hardware diagram of the network topology architecture of the relevant technology intelligent computing center; Figure 4 for Figure 2 A hardware schematic diagram of the illustrated embodiment; Figure 5 The figure is a flow chart of an embodiment of a light path inspection method of the present application. DETAILED DESCRIPTION

[0016] The present application is further described below in conjunction with the accompanying drawings, and the structure and principle of the present application are very clear to people in the field. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0017] Figure 1 It is a schematic diagram of the network topology architecture of the related technology intelligent computing center. It can be understood that in an intelligent computing center, a typical two-layer network topology, namely the Spine-Leaf architecture, the spine switch is located at the top of the network architecture. As a core layer device, it connects all the leaf switches. The spine switch provides high-bandwidth and low-latency connections to ensure that data can be transmitted quickly and efficiently in the network. The leaf switch is located on the second layer and connects the spine switch and the server on the rack. The leaf switch is responsible for the network connection between servers on multiple server racks. The server rack contains one or more servers and is the terminal node of the network. Each server contains several computing power computing units (such as GPUs). The above-mentioned two-layer network connects all the computing power computing units of all servers for parallel computing and communication.

[0018] The servers equipped with computing units are interconnected through the above-mentioned two-layer network topology to perform tasks such as data processing and storage. Each rack of the server rack contains multiple servers for data processing and storage. Network rack: It is used to place network equipment and switches and is the core of network connection.

[0019] The data traffic starts from the computing unit on the server. If the computing unit at the other end of the communication is connected to the same leaf switch, the traffic is forwarded by the leaf switch and reaches the target computing unit of the target server, completing the entire data communication process. If the computing unit at the other end of the communication is connected to a different leaf switch, the traffic is forwarded by the leaf switch to the spine switch, which is forwarded by the spine switch to the leaf switch corresponding to the relevant computing unit of the target server, and then reaches the target computing unit of the target server, completing the entire data communication process.

[0020] Due to its specific working scenario and technical architecture, the intelligent computing center is different from the traditional data center. Under normal working conditions, all computing units of all servers perform parallel collaborative computing and communication for the same computing task. When any link or unit fails, the computing task performed by the entire cluster will be interrupted. Therefore, its requirements for operation and maintenance (including fault diagnosis, fault convergence, fault recovery, etc.) are higher than those of traditional data centers. To this end, it is necessary to adopt technical means different from traditional ones (such as Telemetry, etc.) to support more effective operation and maintenance. In order to support more effective operation and maintenance, this application solves one of the most critical technical problems, which is the collection of low-cost optical network data.

[0021] For this scenario, the following key requirements exist: All-optical network acquisition, the main acquisition object is the optical fiber link; High-resolution acquisition is required. Low-resolution acquisition makes it difficult to capture microscopic details in the network, and thus cannot perceive potential critical issues such as micro-bursts, millisecond or even microsecond jitter, etc. In the intelligent computing center, due to the large number of GPUs or similar computing units, the corresponding number of optical network ports is also large. If all ports are fully acquired, the cost will be huge, and it is impossible to establish a completely 1:1 twin for the physical network.

[0022] Figure 2 is a schematic block diagram of an embodiment of an optical path inspection system of the present application, Figure 3 This is a partial hardware diagram of the network topology architecture of the relevant technology intelligent computing center. Figure 4 for Figure 2 The hardware schematic diagram of the embodiment shown. Figure 2 , Figure 3 and Figure 4 Understandably, The optical path inspection system may include an optical switch (OCS / OSS), a leaf switch, N optical fiber links, a controller, and a collection system platform. A server (Server) is provided on the optical fiber link. The server is used to transmit collection information to the leaf switch. In this embodiment, the server is a computing device equipped with a GPU or other computing power unit. The optical switch is provided with N full-volume collection ports (An), N forwarding ports (Bn), and an output port (Cn). The full-volume collection port is connected to the optical fiber links of all servers in the network and is used for full-volume collection of collection information. The forwarding port is connected to the leaf switch and is used to forward the collection information to the leaf switch. The output port is connected to the collection system platform and is used to output the inspected collection information to the collection system platform. The controller is used to convert the inspection plan of the collection system platform into the configuration rules for the optical switch, and monitor the status information of the optical switch, so as to notify the collection system platform when it can start collecting signals.

[0023] In this embodiment, "full collection" specifically refers to the comprehensive and complete information collection of the designated optical fiber links in the optical network within the required time, and the information of any link or any specified period of time can be selected. Specifically, all optical fiber links involved in communication in the optical network, regardless of whether they are currently active or not, can be included in the collection scope. Not only the physical layer information of the optical signal is collected, but also the information of the data link layer, network layer and even the application layer to ensure that the complete network status is obtained. During the collection process, it is necessary to ensure that the information collection of each optical fiber link is continuous, and the information within any period of time can be selected so as to capture the instantaneous changes and micro-bursts in the network. Through "full collection", the optical path inspection solution can provide more accurate and comprehensive network status information, and provide strong support for the efficient operation and maintenance management of the intelligent computing center.

[0024] In this embodiment, the intelligent computing center network architecture exemplified in this application uses 400G optical fiber for high-speed data transmission, improving the transmission efficiency and capacity of the network. Figure 2 The sensing surface in the middle dotted box represents the optical path inspection system and its back-end acquisition system platform. Figure 2 The collection system platform represents the back-end collection system platform of this application, which realizes real-time monitoring and intelligent analysis of network status through network monitoring and management tools. Figure 3 and Figure 4 The switch in the figure represents the leaf switch in the aforementioned intelligent computing center network architecture. Figure 4 The collection system platform represents the back-end collection system platform of this application.

[0025] The optical path inspection system also needs to include a controller. There is a bidirectional port connection between the controller and the acquisition system platform, and a bidirectional port connection between the controller and the optical switch. The acquisition system platform sends the inspection plan to the controller. The controller is used to output configuration rules according to the inspection plan. The optical switch executes the configuration rules. At the same time, the optical switch returns the status information after executing the configuration rules to the controller. The controller decides when to notify the acquisition system platform to start collecting signals based on the status information. The inspection plan can include inspection frequency, inspection duration and inspection object, and the configuration rules include switching inspection links.

[0026] The optical fiber link of the object to be collected is disconnected, and the two ends of the disconnection are connected to the An and Bn ports of the optical switch respectively (similar to connecting the optical switch in series to the original optical fiber link). The Cn port of the optical switch is used to output the collected signal after splitting. After the optical fiber of the optical link to be collected (mainly the optical fiber line between the server computing node and the Leaf switch in the intelligent computing scenario) is connected to the optical switch with splitting function, data is collected and forwarded, and the normal operation of the link of the original network is not affected. The original service link that has been split continues to be connected to the original service port (An and Bn ports) through the optical switch. The optical path after splitting passes through the optical switch, and according to certain rules, the light is dynamically and regularly switched to the corresponding collection port, thereby realizing the optical path inspection and collection. After that, the optical path output by splitting is connected to the back-end collection system platform (such as the "intelligent computing visualization system") to analyze the data from the physical layer to the application layer of the optical signal accordingly.

[0027] Based on the above optical path inspection system, the present application also proposes an optical path inspection method of the optical path inspection system. Figure 5 The figure is a flow chart of an embodiment of a light path inspection method of the present application.

[0028] Optical path inspection methods may include: Step S1: The full collection port collects the collection information of all optical fiber links, and the forwarding port forwards the collection information of all optical fiber links to the leaf switch.

[0029] Step S2: The acquisition system platform sends an inspection plan to the controller. The controller outputs configuration rules to the optical switch according to the inspection plan. The optical switch executes the configuration rules and returns the corresponding status information to the controller after completion. The controller notifies the back-end acquisition system platform that it can start collecting signals. In step S2, the inspection plan includes inspection frequency, inspection duration and inspection object, and the configuration rules include switching inspection links. The acquisition system platform can send the inspection plan to the controller in the form of a predefined rule interface through the control plane network. The controller converts the inspection plan into configuration rules and sends the configuration rules to the optical switch through the management interface.

[0030] Step S3: The output port outputs the collection information of a fiber link (the link currently being inspected) to the collection system platform.

[0031] Step S4: The collection system platform receives the collection information, determines the failed optical fiber link according to the collection information, and controls the optical switch to further collect the collection information of the failed optical fiber link. In this embodiment, the fault may refer to high latency, link congestion, etc. The collection system platform can focus on the problematic link for inspection.

[0032] This method needs to be implemented at the software level. Specifically, 1. The acquisition system platform can send inspection requirements (such as inspection frequency, inspection time, inspection objects, etc.) to the controller in the form of a predefined rule interface through the control plane network, and the controller will convert it into the corresponding optical switch programming command script.

[0033] 2. The controller translates the inspection requirements into the configuration rules of the optical switch and sends them to the optical switch through the management interface according to a certain rule (for example, a new configuration rule is sent every 10 seconds, and each time the configuration rule is sent, it switches to the next optical fiber link). The optical switch executes the configuration rules and completes the corresponding optical path switching action.

[0034] 3. The optical switch returns the status information after executing the configuration rules to the controller, and the controller decides when to notify the back-end acquisition system platform to start collecting signals based on the status information.

[0035] 4. The collection system platform collects the traffic of the corresponding optical fiber service link and performs subsequent analysis of the collection system.

[0036] 5. If there are no new requirements for the acquisition system platform, repeat steps 2. to 4. above.

[0037] 6. If there are new requirements for the acquisition system platform, repeat steps 1. to 4. above.

Claims

1. An optical path inspection system, characterized in that: The optical path inspection system includes an optical switch, a leaf switch, at least one optical fiber link, a controller, and a collection system platform. A server is provided on the optical fiber link, and the server is used to transmit the collected information to the leaf switch; The optical switch is provided with at least one full collection port, at least one forwarding port and an output port. The number of the full collection ports, the forwarding ports and the optical fiber link is the same. The full collection port is connected to the optical fiber link for full collection of the collection information. The forwarding port is connected to the leaf switch for forwarding the collection information to the leaf switch. The output port is connected to the collection system platform for outputting the inspected collection information to the collection system platform.

2. The optical path inspection system according to claim 1, characterized in that: The optical path inspection system also includes a controller. A bidirectional port is connected between the controller and the acquisition system platform. A bidirectional port is connected between the controller and the optical switch. The acquisition system platform sends an inspection plan to the controller. The controller is used to output configuration rules according to the inspection plan. The optical switch executes the configuration rules. At the same time, the optical switch returns status information after executing the configuration rules to the controller. The controller decides when to notify the acquisition system platform to start collecting signals based on the status information.

3. The optical path inspection system according to claim 2, characterized in that: The inspection plan includes inspection frequency, inspection duration and inspection object, and the configuration rule includes switching inspection links.

4. The optical path inspection system according to claim 2, characterized in that: The optical path inspection system also includes a spine switch, which is connected to the output end of the leaf switch and is used to receive the collected information.

5. The optical path inspection system according to claim 2, characterized in that: The server is a computing device equipped with a computing unit, and the computing unit includes a GPU.

6. A light path inspection method of the light path inspection system as claimed in claim 2, characterized in that: The optical path inspection method comprises: The full collection port fully collects the collection information of all the optical fiber links; The forwarding port forwards the collection information of all the optical fiber links to the leaf switch, and the output port outputs the collection information of one optical fiber link to the collection system platform; The acquisition system platform sends an inspection plan to the controller, and the controller outputs configuration rules to the optical switch according to the inspection plan; The optical switch executes the configuration rule and returns status information to the controller; After receiving the status information of the optical switch completing the switching, the controller notifies the acquisition system platform that it can start collecting signals.

7. The optical path inspection method according to claim 6, characterized in that: The optical path inspection method further comprises: The collection system platform collects and receives the collection information, determines the failed optical fiber link according to the collection information, and controls the optical switch to fully collect the collection information of the failed optical fiber link.

8. The optical path inspection method according to claim 6, characterized in that: The inspection plan includes inspection frequency, inspection duration and inspection object, and the configuration rule includes switching inspection links.

9. The optical path inspection method according to claim 8, characterized in that: The optical switch executes the configuration rule, including: The output port outputs the collection information of another optical fiber link to the collection system platform.

10. The optical path inspection method according to claim 8, characterized in that: The acquisition system platform sends an inspection plan to the controller, and the controller outputs configuration rules to the optical switch according to the inspection plan, including: The acquisition system platform can send the inspection plan to the controller in the form of a predefined rule interface through the control plane network, and the controller converts and translates the inspection plan into the configuration rules, and sends the configuration rules to the optical switch through the management interface.

Images