TWI591983B - Path State Return Algorithm Based on Centralized Control Plane - Google Patents

Description

Path state return algorithm based on centralized control plane

The present invention relates to a path state reward algorithm, especially a cost-effective one, which is compatible with other network management applications and network components, and can be combined with other applications without affecting other applications. Complete detection tools, or as a network self-optimization, and a centralized control plane path state return calculation mechanism, able to track the actual path through which the data stream flows, providing more network administrators and network management applications. Efficient analysis and removal of obstacles, effectively reducing the CAPEX (Capital Expenditure) and OPEX (Operating Expenditure_Operating Expenditure) based on the centralized control plane path state return algorithm.

In traditional networks, in addition to forwarding packets, switches and routers need to understand the network topology and decide the path through which packets of the data stream are passed to the destination. As shown in the traditional network architecture diagram of Figure 1, the control plane 11 responsible for determining the routing rules is dispersed among the switches and routers. If the network administrator 13 needs to customize the routing rules, it must be connected to each network component. set up.

In addition, in order to understand the necessary information such as the network topology, the network components of the traditional network need to exchange various messages, such as Link Layer Discovery Protocol (LLDP) packets, by exchanging their respective networks. Road information, synchronized with each other for network topology and state awareness, and then based on pre-defined network protocols or routing rules, each network component decides how to forward received packets.

Problems with traditional networks include but are not limited to:

1. Distributing the control plane in switches and routers with low hardware resources limits the functionality and flexibility of the network.

2. A huge amount of information exchange makes the operation of traditional networks inefficient.

3. Each network component determines the nature of the forwarding behavior, making it difficult for network administrators to effectively detect network status and debug.

Due to its decentralized control plane, traditional networks lack a centralized management mechanism and cannot provide an effective network management tool for network administrators 13 to track network information for debugging. When there is a problem in the network or when you need to obtain network information due to maintenance, the network administrator 13 can only take the log 21 of each switch or router and analyze it, or return it through the remote host or router. The packet indirectly obtains network information, as shown in the schematic diagram of the traditional network debugging mode.

When the network size exceeds a certain level, the number of logs 21 of the switch will be difficult to analyze, and the network administrator 13 cannot immediately obtain the required network information. In addition, since the network manager 13 cannot determine the path of the packet, only the logs 21 of all switches and routers can be analyzed. On the other hand, the type of information that can be indirectly obtained by the reply 22 packet sent back by the remote host or router is limited, and it is difficult to be an effective basis for maintenance. If the end-to-end reply 22 packet cannot be recovered due to network conditions, it is impossible to obtain the necessary path and network information in this way.

The most well-known network debugging tools in traditional networks include ping and traceroute. Ping works by throwing ICMP echo requests to the target host and calculating the packet loss rate and round trip time based on the received response packets. The disadvantage of Ping is that it can get less network information, and it is more difficult to debug the network based on its results. Traceroute can obtain a list of routers that the packet passes through in the IP network. The operation principle is to sequentially drop ICMP request packets with different Time to Live (TTL) values to the network, and the ICMP timeout is recovered through the recovered. The packet gets the information of each router on the path. The disadvantage of Traceroute is that it only gets the information of the router and cannot get complete information on the path, such as the state of the switch and the link. In addition, the traceroute method will generate more packets for the network. If there are n routers in the path, traceroute will cause a total of O(n ² ) forwarding behaviors for the routers on the path.

In 2014, IBM released a software-defined network (Software-Defined) Networking, SDN) architecture implementation of SDN traceroute. The SDN architecture separates the control plane of the switch from the forwarding plane, allowing the network administrator to control the flow table and forwarding behavior of the switches in the SDN network. The SDN traceroute sends a detection packet to the starting switch to be tested, and then through the programmable features of the SDN switch, the detection packet is copied back to the SDN controller for return status after each link, so as to confirm The complete path information that the packet flows through the network.

However, SDN traceroute assumes that it can retain certain fields in the packet header as back. It is impractical to use the detection packet to prevent all components in the network from being used by the network management application. Although developers of SDN traceroute may be able to use their permissions to require all application developers to comply with this specification, it will still result in a decrease in the freedom of the controlled network, which inevitably limits the functionality of other SDN applications. In addition, the number of bits that can be used in the packet header is limited, which may result in multiple detection packets corresponding to the same label. If the detection packet is copied by other applications and the redundant backup is sent to the network, the SDN traceroute may not be able to. Correct measurement path.

China Patent No. 201406114, "Translation of Communication Period Information for Supplying Network Paths", published on February 1, 2014, refers to a system and method that receives a communication period corresponding to a server. SESSION INFORMATION, and translates the communication period information into translated communication period information including a topology layout parameter and a data parameter. The translated communication period information is directed to direct a controller to supply based on the translated communication period information. A network path. The network path is such that it conforms to the topology layout parameters and data parameters. Compared with the detection packet of the present invention by transmitting the simulated data stream to be detected, and recovering the path state report returned when the network component receives the detection packet, the calculation method for accurately tracking the flow of the data flow through the actual path is not the same. .

As disclosed on August 20, 2014, CN103997432, Measuring system and The method for supporting analysis of OpenFlow application performance refers to a measurement system and method that supports quantitative analysis of OpenFlow application performance. It is based on an OpenFlow network consisting of a controller and n switches connected to the controller, n switches accepting OpenFlow control of the controller; controller and n switch expansions The local logging function and the clock synchronization function become the measurement entity centralized control by the measurement server. Its characteristics include: no centralized performance bottleneck, the measurement itself has little interference to the network application; it can comprehensively acquire the data plane and the control plane information; and can obtain the interaction relationship between the control plane and the data plane. It is not the case that the path state return calculation method of the centralized control plane is used, which can track the actual path state of the data flow and help the network administrator and its network management application to analyze and troubleshoot more efficiently.

The invention provides a path state return algorithm based on a centralized control plane, which can be chased Trace the actual path state through which the data stream flows. Compared to ping, the present invention can report more network information to network administrators. Compared with the traceroute, the present invention can find out the actual path of the specified data stream in the controlled network with a small amount of information, and reduce the burden on the network caused by the detection action. Compared with the SDN traceroute, the present invention uses the payload content of the packet as the basis for detecting the packet, and does not limit the use and modification of the packet header by other applications, and can be highly compatible with other network management applications. , get the correct test results.

In view of the inadequacies of the foregoing systems and methods, the inventor of the present invention is innovative and invented. After a long period of painstaking research and painstaking research, he finally succeeded in researching and developing a path state return algorithm based on a centralized control plane.

The invention proposes a path state return algorithm based on a centralized control plane, which can track the actual path state flowing through the data stream, and assist the network administrator and its network management application to analyze and troubleshoot more efficiently. The design of the present invention focuses on compatibility with other network management applications and network components. By filling in the identification information of the detection packet, it can be correctly obtained without restricting the use and modification of the packet header. Test results.

The invention provides a path state report algorithm based on a centralized control plane, which can accurately track data by transmitting a detection packet simulating a data stream to be detected, and then recovering a path state report returned by the network component when receiving the detection packet. The actual path through which the flow flows helps the network administrator to more efficiently identify problem components to troubleshoot. In addition, considering the compatibility issues with other network management applications, the present invention fills the identification information into the payload of the packet and can detect it without restricting other network management applications. Therefore, it can be combined with other network management applications as an important basic tool for future development of other network management programs.

11‧‧‧Control plane

12‧‧‧ Forwarding plane

13‧‧‧Network Manager

21‧‧‧Log

22‧‧‧reply

31‧‧‧Creation phase

32‧‧‧Detection phase

33‧‧‧Return phase

311‧‧‧ Give the data flow information to be tested

312‧‧‧ Simulated actual packet generation detection packet

321‧‧‧Send the test packet to the origin switch

322‧‧‧ Waiting and receiving the switch for immediate return

323‧‧‧The destination switch has been rewarded

324‧‧‧ has reached the time limit

331‧‧‧Return the final result

41‧‧‧Send detection packets to the originating switch

42‧‧‧The starting switch returns the detection packet to the physical server

43‧‧‧ Pass the test packet to the next switch according to the routing rules

44‧‧‧The switch sends the detection packet back to the server to report the status

45‧‧‧ Pass the test packet to the next switch on the path

46‧‧‧ starting point exchanger

1 is a conventional network architecture diagram; FIG. 2A-2B are schematic diagrams of a conventional network debugging manner; FIG. 3 is a flowchart of a path state report algorithm based on a centralized control plane according to the present invention; FIG. 4 is a schematic diagram of a process of transmitting a test packet according to the present invention.

The invention provides a path state report algorithm based on a centralized control plane, which generates a detection packet according to the data flow information to be detected, and then sends a detection packet to the starting switch. By making the network components controlled by the network actively report the results when receiving the detection packet, the actual path through which the specified data stream flows in the network is confirmed, which serves as a basis for determining the debugging of the network environment.

The detection process of the present invention is mainly divided into three phases, namely, a creation phase 31, a detection phase 32 and a reward phase 33, and the flowchart thereof is as shown in FIG.

In the creation phase 31, a detection packet simulating the data stream packet to be detected is first generated as a detection of the actual path through which the data stream flows in the network, which includes giving the data stream information to be detected 311. The steps and simulation of the actual packet generation detect packet 312 steps. First, the data stream information 311 step to be detected is based on the data stream information to be detected, such as source MAC, destination MAC, source IP, and destination IP, etc., which can be used to identify the packet information, and generate a detection packet. Header to ensure that the detection packet will have the same routing behavior as the data stream packet to be detected. Then, the actual packet generation detection packet 312 step is simulated, and the payload of the detection packet is filled in a unique keyword to distinguish the detection packet from other packets having the same header. This method does not restrict the use and modification of the packet header by other applications, and is better compatible with it.

After the detection packet is created, the detection process is performed, and the detection packet is sent to the switch that the network administrator wants to detect the path, and the path detection is performed. The switch includes at least the switch located at the beginning of the path to be detected. And the switch at the end of the path to be detected, the step The step includes transmitting the detection packet to the start switch 321 step, waiting for and receiving the switch immediate return 322 step, the destination switch has reported 323 determining step, and the reached time upper limit 324 determining step. The detecting packet is sent to the starting switch 321 step, and the detecting packet generated in the creation phase is mainly sent to the starting switch; the process of waiting for and receiving the immediate return of the switch 322, mainly waiting for the detecting packet reported by the receiving switch; The device has reported 323 judging steps, mainly determining whether the end switch has been rewarded, and if so, entering the return final result 331 step, otherwise entering the time limit upper limit 324 judging step; the time upper limit 324 judging step has been reached, mainly determining whether the detecting time upper limit has been To, if yes, enter the return final result 331 steps, otherwise return to the wait and receive the switch instant return 322 steps.

The actual operation of the detection phase 32 can be understood by the schematic diagram of the detection packet delivery process as shown in FIG. First, the detection packet generated in the creation phase is first transmitted to the origin switch 46 (step 41). After receiving the detection packet, the switch will actively send the detection packet back to the physical server where the algorithm of the present invention is located (step 42), and then pass the detection packet to the next switch according to the routing rule (step 43) for subsequent detection. . Steps 41 to 42 are for detecting the operation status of the origin switch, and determining whether the origin switch can actually return the detection packet. Subsequent path detection is repeated in accordance with the pattern from step 44 to step 45 to detect the status of each link and switch: the switch sends the detection packet back to the server to report the status (step 44), and then detects the packet. Passed to the next switch on the path (step 45).

At the end of the data stream to be detected, the switch returns the detection packet, or after reaching the upper limit of the detection time, it enters the reward phase, which includes the step of returning the final result 331. At this point, the final result 331 step is returned, and the network manager returns the actual path and path state of the data stream that is pieced together according to the status of the network component return during the detection phase.

The present invention can be applied to several different routing cases, including but not limited to single path, routing, multipath routing, multicast routing, and the like. In addition, the design of the present invention focuses on compatibility with other network management applications and network components, can be combined with other applications, and constitutes a more comprehensive detection tool or self-optimized as a network. Use.

For a single path data stream, the present invention is able to detect the actual path through which it flows. For the data stream that passes through more than one path, such as multi-path and multicast data stream, since the detection packet is also copied into multiple copies as the actual packet, the present invention can still analyze the packet according to the packet returned by each switch. Information such as the full path through which the data stream flows and the total amount of packets.

Since the present invention uses the payload content of the packet as the basis for detecting the packet, compared with the method of using the packet header field, the actual path of the data stream can be provided as the detection of other modules without affecting the degree of other applications. Based on the optimization, several integrated use cases are listed below: By combining with the present invention, the routing application can check whether the actual path of the data stream is consistent with the operation logic after the path is established, as a basis for debugging. The Rerouting application can know whether the current path is unblocked according to the detection result of the present invention, and obtain the delay time of the path return. When the data stream cannot reach the destination through the current path or the delay is too high, the path is automatically replaced. Through the actual path of the data stream returned by the present invention, the Statistics and Stress Test applications will be able to accurately target the actual path through which the data stream flows, thereby avoiding an additional burden on the network.

In addition to detecting network status, the present invention can also be used to detect the operation of a server, such as a DHCP server. By filling the actual packet content sent by the client into the header and payload of the detection packet, and sending the detection packet simulating the actual packet to the connection client. The device can transmit the detection packet to the server via the same path as the actual packet, and then obtain the response packet of the server as the basis for judging the state of the server.

The core of the algorithm of the present invention is to uniformly collect the return of the network component when receiving the detection packet, and fill the identification information into the payload instead of the header to avoid affecting the operation of other applications. The algorithm of the present invention can be implemented as long as the network architecture used can satisfy the above two conditions.

One of the choices for implementing the algorithm of the present invention is the SDN (Software-Defined Networking) network architecture. We can use the algorithm of the present invention as an SDN network management application, which can be programmed by using the SDN network architecture. The characteristics of the custom SDN switch forwarding behavior, so that all SDN switches receive the detection packet, in addition to forwarding according to the original routing rules, and then report the path status information to the application of the invention to achieve centralized control design .

When the path state report algorithm based on the centralized control plane is compared with the conventional technology, the following advantages are obtained: a path state report algorithm based on the centralized control plane can track the flow of data streams. The actual path state.

1. Compared to ping, the present invention can report more network information to network administrators.

2. Compared with the traceroute, the present invention can find out the actual path of the specified data stream in the controlled network with a small amount of information, and reduce the burden on the network caused by the detection action.

3. Compared with SDN traceroute, the present invention uses the payload content of the packet as the basis for detecting the packet, and does not limit the use and modification of the packet header by other applications, and is highly compatible with other network management applications. Under the premise, get the correct test results.

4. The present invention focuses on compatibility with other network management applications and network components, and can be combined with other applications without compromising other applications to form a more complete detection tool or as a network self-optimization. .

5. The present invention adopts a centralized control plane path state return calculation mechanism, which can track the actual path through which the data stream flows, and provides network administrators and network management applications with the ability to analyze and remove obstacles more efficiently, and can greatly Reduced CAPEX and OPEX.

The detailed description of the present invention is intended to be illustrative of a preferred embodiment of the invention, and is not intended to limit the scope of the invention. The patent scope of this case.

To sum up, this case is not only innovative in terms of technical thinking, but also has many of the above-mentioned functions that are not in the traditional methods of the past. It has fully complied with the statutory invention patent requirements of novelty and progressiveness, and applied for it according to law. Approved this invention patent application, in order to invent invention, to the sense of virtue.

31‧‧‧Creation phase

32‧‧‧Detection phase

33‧‧‧Return phase

311‧‧‧ Give the data flow information to be tested

312‧‧‧ Simulated actual packet generation detection packet

321‧‧‧Send the test packet to the origin switch

322‧‧‧ Waiting and receiving the switch for immediate return

323‧‧‧The destination switch has been rewarded

324‧‧‧ has reached the time limit

331‧‧‧Return the final result

Claims (7)

A path state return algorithm based on a centralized control plane, which mainly comprises: a creation phase, which simulates an actual packet to generate a detection packet according to the data stream information to be detected; and a detection phase, sends a detection packet to the desired detection path. The switch, the waiting and receiving switch real-time return, the switch at the end of the desired path has been judged and the time limit has been judged, and the switch at the end of the path to be detected has returned the switch that judges the main end point. Whether it has been reported, if it is, it will enter the return stage, otherwise it will enter the step of the time limit judgment. The time limit judgment step is mainly to judge whether the detection time limit has arrived, and if so, enter the return stage, otherwise return to the waiting and receive switch. The step of immediate return; a return phase that returns the actual path and path state through which the data stream flows. The path state report algorithm based on the centralized control plane according to claim 1, wherein the actual packet is generated according to the data stream information to be detected, and the detection packet is generated according to the data stream information to be detected. Header to ensure that the detection packet will have the same routing behavior as the data stream packet to be detected. The centralized state plane-based path state report algorithm according to claim 1 or 2, wherein the data stream information to be detected includes information of a source MAC, a destination MAC, a source IP, or a destination IP that can be used to identify the packet. The path state report algorithm based on the centralized control plane, as described in claim 1, wherein the actual packet generation detection packet is mainly used to detect the payload of the packet to fill in a unique keyword, so that the detection packet has the same header as the other. The package is different. The path state report algorithm based on the centralized control plane according to claim 1, wherein the detection packet is sent to the switch located at the start of the path to be detected, and the switch is generated at the creation stage. . The centralized state plane based path state report algorithm as claimed in claim 1, wherein the waiting and receiving switch immediate return is mainly a detection packet reported by the switch waiting for the receiving end. The path state report algorithm based on the centralized control plane according to claim 1, wherein the reward phase is the actual path and path state of the data stream that is patched according to the status of the network component report when the network manager reports the detection phase. .