CN112866120A

CN112866120A - SDN flow table loop-free consistency updating method and system based on classified search

Info

Publication number: CN112866120A
Application number: CN202110114292.1A
Authority: CN
Inventors: 张奇支; 杨荣宽; 郑伟平; 赵淦森; 庄捷; 钟剑豪
Original assignee: South China Normal University
Current assignee: South China Normal University
Priority date: 2021-01-26
Filing date: 2021-01-26
Publication date: 2021-05-28

Abstract

The embodiment of the invention relates to the technical field of computer networks, in particular to a novel SDN network architecture. Specifically, the system designs a switch classification model and a loop search optimization model. And in the updating process, the classification model is used for analyzing the states of the switch before and after updating and initializing. And then, calling a loop detection module in the loop search optimization model to detect the nodes in the switch node set to be updated, and acquiring the updatable nodes in the first round for updating. And then, circularly calling a searching module of the optimization model in the non-updated node set to search for the updatable nodes meeting the conditions until the updating is completed. A new switch classification model and a new search model are constructed by improving a classification mode in a classification time sequence updating algorithm, the applicability of an application scene can be increased, the updating operation complexity of a switch can be reduced, the updating waiting time delay can be reduced, and the updating efficiency of an SDN flow table can be improved.

Description

SDN flow table loop-free consistency updating method and system based on classified search

Technical Field

The embodiment of the invention relates to the technical field of novel network architectures of computers, in particular to a method and a system for updating loop-free consistency of an SDN flow table based on classified search.

Background

SDN (Software Defined Networking) is a new network architecture with programmability that separates a network control plane and a data plane. The SDN architecture based on the OpenFlow protocol abstracts out the underlying infrastructure, enabling users to centrally control network forwarding behavior through software programming. When the network state changes, the SDN controller may update the switch flow table entry by issuing a new forwarding rule, thereby managing the entire network state. Although the SDN realizes centralized control of a network, the distributed characteristics of the data plane switch also bring many problems, such as speed of installing flow entries by the switch itself, transmission delay between switches, and the like, so that the effective time of the flow entries is difficult to be consistent, resulting in inconsistent update, and increasing difficulty of centralized management.

In an SDN network, when a switch receives a datagram, it first looks up a forwarding flow table. And if the matched flow table items exist, executing corresponding actions, otherwise, sending packetin information to the controller, and reporting the message to the controller for processing. After the controller processes the forwarding flow, the controller issues a new forwarding flow entry to the switch to guide subsequent forwarding. In addition, when the network state changes, the controller actively or passively sends a new forwarding flow table entry to the switch to form a new forwarding path.

Because the data plane switches are arranged in a distributed manner, the time delay of the controller for issuing a new flow table to each switch is different, and the updating speed inside the switches is also different, which can cause that all flow table items of a new rule cannot take effect at the same time, so that the data packet is subjected to wrong forwarding processing, problems of data forwarding interruption, loops and the like are caused, and the reliability of network flow transmission and the guarantee of fine-grained QoS are influenced.

Therefore, during the update of the old and new rules, it is necessary to keep the consistency update of the flow table. That is, when the controller updates the flow entries in the switches, the packets being transmitted either match the old rule flow entries or the new rule flow entries at each switch, and the case where the old rule is matched at switch a and the new rule is matched at switch B cannot occur. To better illustrate the problem caused by the update inconsistency of the SDN flow table, as shown in fig. 1, where a node Ni (1 ≦ i ≦ 8) represents a switch, a solid line represents an old forwarding path, and a dotted line represents a new forwarding path.

Suppose that the SDN controller issues update commands to each switch at the same time, and updates the forwarding rule to a new path. Due to the difference in transmission delays between the controller and the different switches, the switches cannot complete the update simultaneously. When N1 completes the update and N8 has not, i.e. N8 does not have a forwarding rule matching the data flow, the data flow forwarded to N8 will not know what to do, which would cause a flow interruption if the default action was to drop the message. In addition, if the switch N6 is updated and N5 is not updated, a loop N6- > N5- > N6 is formed, and the link is occupied for a long time until the TTL of the packet is over time, which may reduce the link utilization rate and the forwarding performance.

Generally speaking, the problem of inconsistent update of the SDN flow table may cause network interruption, loop, security hole, and the like. To ensure that the network can forward the packets correctly during the transition from the old rule to the new rule, all intermediate states that may occur and their attributes must be considered. Obviously, the consistency of the flow table is updated, and whether the network configuration can be correctly realized is related.

The prior art provides a flow table updating consistency scheme based on classification and classification time sequence, and the core lies in classifying switches according to updating operation types and updating rules according to a certain sequence. In contrast, the classification timing update adds a subdivision operation to the flow table entry. Although the algorithm of the classification time sequence updating is simple, the nodes on the new forwarding path and the old forwarding path are updated according to the strictly set sequence, the condition that the nodes are simultaneously switches on the new forwarding path and the old forwarding path is ignored, and the application scene limitation exists. In addition, regardless of the complexity of the updating topology, the next stage can be performed only by waiting for a certain end-to-end longest time delay, the updating round of the scheme is fixed, and the updating time delay is too long.

Based on the node dependency relationship, a reverse update scheme is also proposed in the prior art. The scheme recursively updates forward from the destination node according to the new rule forwarding path to avoid loops. Although the algorithm is simple, the node dependence is complex, the condition that the node rule does not need to be updated is not considered, and the forwarding path is updated from the destination node in a recursion forward mode strictly according to the new rule so as to avoid a loop. Therefore, the number of update rounds is large, so that the update time is prolonged. In addition, based on the reverse update, a bidirectional update scheme is also proposed. And searching for nodes meeting the two conditions that all child nodes are updated to the new rule in the current updating state or all parent nodes are updated to the new rule in the final state. However, at most 2 nodes are updated each time, the number of update rounds is still excessive. Aiming at the dependence problem and optimization updating, each node is set to have 3 states, and under the condition of ensuring no ring, as many nodes as possible are updated in parallel, so that the updating round is reduced to the maximum extent. However, the scheme needs global traversal to update the node state in each round, and the complexity of node search is high.

Disclosure of Invention

The invention aims to provide a method and a system for updating the loop-free consistency of an SDN flow table based on classification search, which improve the classification mode in a classification time sequence updating algorithm, construct a new switch classification model, and construct a corresponding loop search optimization model by analyzing the difference of forwarding rules before and after updating based on a classified node set, thereby optimizing the complexity and the updating efficiency of updated search calculation, increasing the applicability of an application scene and reducing the waiting time delay of updating.

To solve the above technical problem, in a first aspect, an embodiment of the present invention provides a method for updating loop-free consistency of an SDN flow table based on classified search, as shown in fig. 2, including:

step S1, obtaining a first to-be-updated in the network configurationNode set S_nodeSecond node set S without updating_unThird node set S only needing to delete old rules_{del_old}Fourth node set T that is updateable per round_iAnd an updated fifth set of nodes S_updated；

Step S2, detecting the first node set S one by one_nodeAdding nodes which do not introduce a loop to the fourth node set T_i；

Step S3, updating the third node set S in parallel_{del_old}And a fourth node set T_iA node of (2);

step S4, searching a first node set S_nodeAnd a fifth set of nodes S_updatedTo add to the fourth set of nodes T_iUpdating the fourth node set T_iUntil the first set of nodes S_nodeAnd a fifth set of nodes S_updatedThe difference set of (c) is null.

In a second aspect, an embodiment of the present invention further provides an SDN flow table loop-free consistency updating system based on classified search, including:

a classification module for obtaining a first node set S to be updated in the network configuration_nodeSecond node set S without updating_unThird node set S only needing to delete old rules_{del_old}Fourth node set T that is updateable per round_iAnd an updated fifth set of nodes S_updated；

A loop detection module for detecting the first node set S one by one_nodeAdding nodes which do not introduce a loop to the fourth node set T_i；

An update module for updating the third node set S in parallel_{del_old}And a fourth node set T_iA node of (2);

a search module for searching the first set of nodes S_nodeAnd a fifth set of nodes S_updatedTo add to the fourth set of nodes T_iUpdating the fourth node set T_iUntil the first set of nodes S_nodeAnd a fifth nodeCollection S_updatedThe difference set of (c) is null.

In a third aspect, an embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor, when executing the program, implements the steps of the SDN flow table loop-free consistency updating method based on classified search according to the embodiment of the first aspect of the present invention.

In a fourth aspect, an embodiment of the present invention provides a non-transitory computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the steps of the SDN flow table loop-free consistency updating method based on classification search according to an embodiment of the first aspect of the present invention.

Compared with the prior art, the method and the device have the advantages that the classification model is used for analyzing the states of all the nodes before and after updating, and initializing the nodes; and then, calling a loop detection algorithm to detect the nodes of the corresponding node set, and acquiring the updatable nodes for updating. And finally, searching for the updatable nodes meeting the conditions by adopting a search algorithm in a centralized and cyclic manner in the non-updated nodes until the updating is completed, improving the classification mode in the classification time sequence updating algorithm, and constructing a new switch classification model which can increase the applicability of the application scene.

In addition, the step S2 includes:

one-by-one detection of first node set S based on topological sorting algorithm_nodeThe node in (1) is directly the first node set S_nodeAdding a new rule to the node in the node, and acquiring check _ loop (N) if the node is judged to be known_i,P_old,P_new) When True, node N_iWill not introduce a loop, will connect the node N_iAdd to fourth node set T_iI is 1; wherein, P_oldIndicating the original transmission path of the data in the network configuration before updating, P_newThe final transmission path of data, check _ loop (N), representing the network configuration after the next update_i,P_old,P_new) For judging node N_iFrom P_oldUpdate transition to P_newWhether a loop will be created.

In addition, in step S3, the fourth node set T is updated_iAfter the node, further comprising:

for the fourth node set T_iAfter issuing the old rule corresponding to the new rule replacement node, the node is transferred to a fifth node set S_updatedIn (1).

In addition, in step S3, the third node set S is updated_{del_old}And then, the method further comprises the following steps:

for the third node set S_{del_old}And (5) directly deleting the old rule corresponding to the node by the node in the system, and then removing the node.

In addition, in the step S4, the first node set S is searched_nodeAnd a fifth set of nodes S_updatedTo add to the fourth set of nodes T_iThe method specifically comprises the following steps:

searching N in sequence_i＝S_node－S_updatedCan update the node N in_iIf it is judged to obtain N_iIf no father node or the father nodes are updated, the node N is connected_iJoin to fourth node set T_iIn the sum, i is more than 1 and less than or equal to r_max，r_maxRepresenting the maximum number of update rounds required for the entire update process.

Further, in step S1:

wherein, nexthop (s, P)_old/P_new) Representing node s on path P_oldOr P_newNext hop of (s, P)_old,P_new) True/False indicates that updating the node yes/no will generate a loop; length refers to the number of nodes involved in the transmission path, S_subiRepresenting the node set to be updated, r, remaining after each round of updating_maxRepresents the maximum number of update rounds, judge _ allgather (s, P), required by the entire update process_old,P_new) For determining whether all of the parent nodes of node s have been updated to the new rule.

Drawings

One or more embodiments are illustrated by way of example in the accompanying drawings, which correspond to the figures in which like reference numerals refer to similar elements and which are not to scale unless otherwise specified.

FIG. 1 is a flow diagram of an example prior art route update;

fig. 2 is a flowchart illustration of a method for updating loop-free consistency of an SDN flow table based on classified search according to a first embodiment of the present invention;

FIG. 3 is an update flow of a CSCU scheme according to an embodiment of the invention;

FIG. 4 is a diagram illustrating an example of an update process for a CSCU scheme, according to an embodiment of the invention;

FIG. 5 is a diagram illustrating update rounds according to various aspects of an embodiment of the present invention;

FIG. 6 is a diagram illustrating update turn cases of CSCU schemes at different levels of interleaving according to an embodiment of the present invention;

FIG. 7 is a comparison of classification timing updates at different levels of interleaving in accordance with an embodiment of the present invention;

FIG. 8 is a schematic diagram illustrating operational complexity of an update node for different path hops, according to an embodiment of the present invention;

fig. 9 is a schematic diagram of a server according to a third embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention more apparent, embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, it will be appreciated by those of ordinary skill in the art that numerous technical details are set forth in order to provide a better understanding of the present application in various embodiments of the present invention. However, the technical solution claimed in the present application can be implemented without these technical details and various changes and modifications based on the following embodiments. The following embodiments are divided for convenience of description, and should not constitute any limitation to the specific implementation manner of the present invention, and the embodiments may be mutually incorporated and referred to without contradiction.

The terms "first" and "second" in the embodiments of the present application are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present application, the terms "comprise" and "have", as well as any variations thereof, are intended to cover a non-exclusive inclusion. For example, a system, product or apparatus that comprises a list of elements or components is not limited to only those elements or components but may alternatively include other elements or components not expressly listed or inherent to such product or apparatus. In the description of the present application, "plurality" means at least two, e.g., two, three, etc., unless explicitly specifically limited otherwise.

Aiming at the problems of poor universality, prolonged updating time, high calculation complexity of optimized updating and the like of the classification time sequence updating application scene, the embodiment of the invention provides a ring-free updating Consistency Scheme (CSCU) based on classification searching on the basis of classification updating and optimized updating. In the CSCU scheme, the embodiment of the invention designs a switch classification model, and designs a loop search optimization model by combining a node dependence idea on the classification basis, wherein the optimization model comprises a loop detection module and a search module. The update process of the combination of the two models can be summarized as follows: firstly, analyzing the states of all nodes before and after updating by using a classification model, and initializing the nodes; and then calling a loop detection module of the loop search optimization model to obtain the updatable nodes of the corresponding node set for updating. And finally, circularly calling a searching module of the optimization model in the non-updated node set to search for the updatable nodes meeting the conditions until the updating is completed. The following description and description will proceed with reference being made to various embodiments.

A first embodiment of the present invention relates to a method for updating loop-free consistency of an SDN flow table based on classified search, as shown in fig. 2, including:

s1, acquiring a first node set S to be updated in network configuration_nodeSecond node set S without updating_unThird node set S only needing to delete old rules_{del_old}Fourth node set T that is updateable per round_iAnd an updated fifth set of nodes S_updated；

S2, detecting the first node set S one by one_nodeAdding nodes which do not introduce a loop to the fourth node set T_i；

S3, and updating the third node set S in parallel_{del_old}And a fourth node set T_iA node of (2);

s4, searching the first node set S_nodeAnd a fifth set of nodes S_updatedTo add to the fourth set of nodes T_iUpdating the fourth node set T_iUntil the first set of nodes S_nodeAnd a fifth set of nodes S_updatedThe difference set of (c) is null.

Updating an SDN flow table: when the network configuration is changed, a new forwarding rule (rule is a flow table) is issued by the SDN controller in a unified manner to update the rule of the switch, so that the switch is guided to forward. But when changing the network configuration to improve network performance, because the switches asynchronously update behavior: factors such as an update sequence of network device nodes and interaction delay among devices may affect an installation sequence of an SDN controller issuing a new rule to a switch, thereby affecting update consistency of an SDN flow table and final correctness of data forwarding. The embodiment of the invention provides a feasible and efficient updating method to ensure that updating of an SDN flow table is consistent

Specifically, in this embodiment, a directed graph is used to represent a point-to-point communication network, and nodes represent switches. The update of the network configuration may be defined as:

wherein E_srcRepresenting the network configuration before updating, E_dstRepresenting the updated network configuration, U being represented by E_srcTransition to E_dstThe set of update operations. P_oldAnd P_newEach represents E_srcOriginal transmission path of data and E_dstThe data of the following finally transmit the route. Meanwhile, nexthop (s, P) is defined_old/P_new) Representing node s on path P_oldOr P_newFor the next hop, use

Represents E_srcUpdate transition to E_dstWhether a loop will be created. It is known that, during the updating process, the configuration is updated or not successfully by the real-time interaction time delay D (P) between the devices_old→P_new) Real-time update operation of a device U (P)_old→P_new) As a result, problems such as loops in the network may occur due to the above factors. Based on the above definitions, the problem of the scheme is described as: known as E_srcAnd E_dstBy analysis of P_oldAnd P_newIn the presence of

Under the conditions of (1), D (P) is sought_old→P_new) Less U (P) with better updating performance_old→P_new) To ensure

And (4) correctly transiting.

The purpose of this embodiment is to

In the process, an updating sequence which has less time delay and is superior to the node searching method can be updated. In the process of constructing the model, the embodiment of the present invention needs to create and update the following related sets based on the concepts of classification and node-dependent search, as shown in table 1. In this embodiment, a node classification method is constructed, classification is performed through a node classification model, and the following relevant set information needs to be maintained in the classification model construction process, where length refers to the number of nodes and S involved in a transmission path_subiRepresenting the node set r to be updated left after each round of updating_maxRepresents the maximum number of update rounds, check _ loop (s, P), required by the entire update process_old,P_new) For judging whether the update node s will introduce a loop, judge _ allfane (s, P)_old,P_new) For determining whether the parent nodes of node s have all been updated to the new rule.

TABLE 1 set of related nodes

Network configuration before update E_srcThe original transmission path of the data is P_oldUpdated network configuration E_dstThe final transmission path of the data is P_new. The SDN controller may analyze and compare nodes (taking a node s as an example below) involved in the two paths according to the two data transmission paths before and after the update: there are several situations as follows:

1. node s is at P_oldBut not at P_newIn this case, the node does not include the node in the transmission path of the packet after the network configuration is updated, and therefore the node only needs to perform the update operation of deleting the old rule. Adding into node set S_{del_old}In (1).

2. Node s is in Pold, also in P_newWhile nexthop (s, P)_old) Is equal to nexthop (s, P)_new) I.e. node s is on the old wayThe next hop of the path and the next hop of the node s in the new path are the same node. The rule of the node s before and after updating is the same, and the node s does not need to be updated and does not need to execute updating operation. Adding an entry node S_unIn (1).

3. Node s is at P_oldIs also at P_newMedium, but nexthop (s, P)_old) Is not equal to nexthop (s, P)_new) I.e. the next hop for node s on the old path is different from the next hop for node s on the new path. And the node s is required to execute an updating operation of installing a new rule to replace an old rule. Adding an entry node S_nodeIn (1).

In addition, due to T_iIs to store the updateable nodes found in each round, S_updatedIs to store the updated nodes, but not to search and update in the initialization stage, so T is initialized_i＝null，S_updated＝null。

The definition of the node dependency relationship is given first: when node Na must be updated after the update of node Nb is completed, node Na depends on node Nb. In addition, since the destination node is not involved in rule updating, it is defined that no update node is required. The updating flow of the CSCU scheme is shown in fig. 3.

The embodiment also provides a loop search optimization model: after the model acts on the classification model, the updatable nodes in each round are searched for updating, and the model comprises two modules:

1. a loop detection module:

the module is mainly used for sorting S_nodeAnd selecting nodes which can be updated in the first round in the node set. The method specifically comprises the following steps: to S_nodeThe nodes in the node set are detected one by one, namely, a new forwarding rule is directly installed for the nodes, and then whether a loop is caused or not is judged.

In fact, the path updated to the final state by default is loop-free, i.e. the old path is loop-free. After a new forwarding rule is installed for a node s, only the node s needs to be judgedWhether there is a path back to s at the next hop s _ next of the new rule. The detection method based on the topological sorting algorithm can judge whether a loop returning to the node per se exists, so that the node found by the loop detection module does not generate the loop and can be updated. To S_nodeThe nodes in the node set are detected one by one, namely, a new forwarding rule is directly installed for the nodes, and if no loop is generated, the nodes are added into the node set T_i(in this case, i is 1). If yes, skipping directly and continuing to detect the next node. Note that after each node is detected, the new rule installed for detection is deleted and the next node is detected. After the detection is finished, the node set T is updated in parallel_iNode set S obtained by classifying with classification model_{del_old}The node (2) is as follows: wherein, for T_iThe node in (2) performs the following operations: issuing a new rule to replace the old rule, and transferring the node to a node set S_updatedIn (1). For S_{del_old}And (4) directly deleting the corresponding old rule from the node in the database, and then removing the node.

First, the CSCU scheme utilizes a node classification model pair E_srcAnd E_dstMaking a comparison by comparing P_oldAnd P_newNode state, initialize the node set in table 1.

Then, calling a loop detection module based on a topological sorting algorithm to detect S one by one_nodeThe nodes in (1) add new rules to the nodes one by one, if check _ loop (N)_i,P_old,P_new) True, i.e. no loop is introduced, N will be_iAdding to node set T_i(in this case, i is 1). After the detection is finished, the node set T is updated in parallel_iAnd S_{del_old}Of the node (b). Wherein, for T_iThe node in (2) performs the following operations: issuing a new rule to replace the old rule, and transferring the node to a node set S_updatedIn (1). For S_{del_old}And (4) directly deleting the corresponding old rule from the node in the database, and then removing the node. Whether the old rule is deleted is also the basis for judging whether one node is updated. S_{del_old}Storing nodes that only need to delete old rules, since the node rule failure time is uncertainIf S is not deleted in time_{del_old}The old rule of the middle node can influence the algorithm of the scheme for searching the updatable node according to the dependency relationship. When the above node set update is completed, the algorithm continues.

Subsequently, the search module pair N is called_i∈S_node-S_updatedSearching for the updatable nodes in turn: will satisfy the condition Find _ false (N)_i) Null or judge _ allfanther (N)_i,P_old,P_new) True node N_iIs added to T_iIn (1), the first two conditions represent N_iNo parent node or the parent node is updated. After the search is finished, T is updated in parallel_iNode of, simultaneously with T_iNode transfer to node set S_updatedThe algorithm continues.

For node set S_node-S_updatedThe search module is called by circulating execution until S_node-S_{updated＝null}And the scheme ends. The following description of this embodiment will prove that, in combination with a specific classification algorithm and a specific search algorithm, the CSCU scheme can effectively improve the update efficiency and has lower computational complexity.

Taking the topology shown in fig. 4 as an example, the updating process of the CSCU scheme is demonstrated by abstracting the network into a point-to-point directed graph. Wherein the solid lines represent the pre-update data forwarding paths of the network configuration and the dashed lines represent the post-update data forwarding paths. The updating steps are as follows:

s obtained after the classification model is completed_node＝{N₁,N₂,N₃,N₄,N₆,N₇,N₉,N₁₁}. And detecting the nodes in the node set one by one: first, supply N₁Installing a new rule;

N₁the next hop in the new rule is N₃So we only need to judge N₃Whether or not there is a return to N₁The path of (A) is, as can be seen in FIG. 4, N₃Only N₃->N₄->...N₁₀Transmission path of (3), without returning to N₁So to give N₁Installing new rules directly does not cause loops, so N can be used₁Adding into T₁In (1).

In the same way, for S_nodeThe other nodes can obtain: directly to node N₂,N₄,N₆,N₁₁No loop is introduced by adding new rule detection, so T is obtained₁＝{N₁,N₂,N₄,N₆,N₁₁}。

However, for S_nodeNode N of₃,N₇,N₉Detecting, and respectively and directly installing new rule detection;

it is clear that N is detected₃Will cause a loop N₂->N₃->N₂(ii) a Detecting N₇Will cause a loop N₆->N₇->N₆(ii) a Detecting N₉Will cause a loop N₇->N₈->N₉->N₇. Therefore, the three nodes remain in the original state.

Node set T is then updated in parallel₁＝{N₁,N₂,N₄,N₆,N₁₁And node set S_{del_old}＝{N₅Nodes in (1): install new rule to replace old rule, and replace T with T₁Node of (2) is transferred to the updated node set S_updatedIn (1). As shown after the first round of updating: wherein the dotted line represents the node forwarding path (node N) after the round of updating₈No update is required, where the default is already updated to facilitate distinguishing whether an update is complete).

The search module is mainly used for carrying out subsequent updating after the loop detection module is finished: and under the current state, selecting the nodes meeting the updating condition from the non-updated nodes for updating. S_node-S_updatedThe set difference set of (A) represents the set of nodes that have not been updated, only if S_node-S_updatedWhen null, the updating is completed, otherwise, the searching updating is circularly carried out. The nodes satisfying the update condition refer to: all parents of the current node are updated or the current node has no parents. Definition of parent-child nodes: if Na->Nb->Nc, Na is Nb, Nc parent node, Nb is Nc parent node, and Nc is Na and NbAnd (4) child nodes. Conversion to a representation of a directed graph: for Na, if a node has a path to reach Na, the node is the parent node of Na. "C (B)

After the loop detection module is completed, S_node-S_updated＝{N₃,N₇,N₉And judging whether the node meets the updating condition or not in the current state. At the present state, N₃Is N₁And N is₁Is updated, so that the update condition is satisfied, and is added to T₂Performing the following steps; n in the current state₇Without father node, satisfying updating condition, adding to T₂Performing the following steps; n is a radical of₉Has a parent node of { N₈,N₁₁,N₄,N₂,N₃,N₁In which N is₃Is only currently updatable, but not yet, so N₉The update condition is not satisfied, so the original state is maintained. Therefore, the searched updateable nodes are: t is₂＝{N₃,N₇}. And executing an updating operation: install new rule to replace old rule, and replace T with T₂Node of (2) is transferred to the updated node set S_updatedIn (1). Obtaining a graph 4: where the dashed lines represent the new paths after having been updated and the solid lines represent the old paths.

Because of S_node-S_updatedNot null, so search updates are cycled through, at which time S_node-S_updated＝{N₉}. To S_node-S_updatedThe node(s) of (1) determines whether the update condition is satisfied, at which time N₉Has a parent node of { N₈,N₁₁,N₄,N₂,N₃,N₁All have been updated, so N₉Meets the updating condition and is added into T₃＝{N₉}. And executing an updating operation: install new rule to replace old rule, and replace T with T₃Node of (2) is transferred to the updated node set S_updatedIn (1). Obtaining a graph 4: where the dashed lines represent the new paths after having been updated and the solid lines represent the old paths. At this time, S_node-S_updatedNull, the update is complete.

In this embodiment, the CSCU scheme is compared with other schemes by taking fig. 4 as an example:

1) updating based on the classification time sequence: the controller initializes the switches, and the switches are divided into four types: ingress switches, new path switches, old path switches, egress switches. ② update the export exchanger { N₁₀And new path switch S_new＝{N₁,N₃,N₂,N₄,N₁₁,N₈,N₉,N₇,N₆And executing operation of inserting the flow table entry to be newly added. Thirdly, after waiting for a longest end-to-end time delay, updating the ingress switch { N }₁And executing the operation of modifying the flow table entry. Fourthly, waiting for the longest end-to-end time delay and updating the old path switch S_old＝{N₁,N₂,N₃,N₄,N₅,N₆,N₇,N₈,N₉And egress switch { N }₁₀And executing the operation of deleting the flow table items, and finishing all updating. Obviously, when the switches involved in the old and new rules have complex associations, the scheme needs to perform different flow table operations on a large number of common switches and switches that do not need to be updated at different stages, so that the number of times of update operations of the switches is increased, which may increase the waiting time for the update.

2) And (3) reverse updating: all nodes of the scheme are recursively updated forward from the destination node strictly according to the new rule. Since the default destination node does not need to be updated, the required update round is the number of nodes-1 involved in the new rule. As can be seen from the example of fig. 3, the number of nodes involved in the forwarding path of the new rule is 10, i.e., 9 rounds are required to complete the update.

3) Bidirectional updating: according to two search conditions to obtain { (N)₁,N₆)->(N₂,N₇)->(N₃,N₉)->(N₄,N₈)->(N₅,N₁₁) And 5 rounds are needed to complete the updating.

4) And (3) optimizing and updating: sequentially adding new rules to perform loop-free detection on all nodes, wherein the found dependent forest root nodes have { N }₁,N₂,N₄,N₅,N₆,N₁₁And deleting the old rules respectively, and searching the child nodes through loop-free detection to obtain the dependency tree: n is a radical of₂->N₃,N₆->N₇->N₉So the whole dependent forest is { N₁,N₂->N₃,N₄,N₅,N₆->N₇->N₉,N₁₁}. Wherein, the maximum height of the dependency tree in the dependency forest is the required updating turn, namely 3 turns are required for completing the updating.

The CSCU proposal provided by the embodiment of the invention is that after the classification model is executed, the T is obtained by executing a loop-free detection module of a loop search optimization model₁＝{N₁,N₂,N₄,N₆,N₁₁Get T by the recycle execution search module₂＝{N₃，N₇}，T₃＝{N₉}. Therefore, the updating can be completed through 3 rounds.

In this embodiment, since the acyclic property is a main research point of the scheme, the scheme algorithm acyclic property verification is also performed:

1) first, no loop is introduced by the loop detection module in the loop search optimization model. In fact, during the updating process, the loop-free detection module based on the topological sorting is used for the node set S_nodeThe node in (1) detects according to whether a loop is generated by directly adding a new rule. If a loop is generated, the state is kept unchanged, otherwise, the loop is added into the node set T one by one_iIn (1). Then updating T in parallel_iAnd S_{del_old}The node (2) is as follows: for s ∈ T₁Performing old and new rule replacement and joining S_updatedIn the pair S ∈ S_{del_old}Old rule deletion is performed. Parallel updating T because nodes determine whether to update only through loop-free detection_iThe process of the node in (1) obviously does not introduce a loop. In addition, for S ∈ S_{del_old}And executing the operation of deleting the old rule, only releasing the memory space of the corresponding switch, and not causing loop-free. Therefore, the loop-free detection update procedure does not introduce loops.

2) Secondly, the loop search module in the loop search optimization model does not introduce loops. In fact, the path updated to the final state by default is loop-free. Now assume the current stateWith a local link N₁->N₂->N_cur，N_curNot yet updated, its parent node N₁，N₂Are updated. Updating N_curForming a loop N_cur->N₁->N₂->N_curAnd, according to the assumption, its parent node is updated, i.e. the nodes involved in the entire loop link are updated to the new rule. The loop is a local link updated to the final state entirely, and there is no contradiction between the loop and the new rule in the final state, so update N_curNo loops are introduced. Searching and updating are circulated until S_node-S_updatedWhen null, the updating is finished, and no loop appears in the whole updating process.

The embodiment also compares 4 aspects of applicability, updating time, complexity of node operation updating and complexity of algorithm calculation updating. And (4) universality. The applicability and the application scene range of the scheme are indicated. And updating the time. The total time used in each stage of the whole updating process is centrally embodied in the required round of the updating process, and the more rounds are required for updating, the longer the time for waiting for updating is, and the longer the final total updating time is. And updating the operation complexity of the node. The more nodes that need to execute different update operations in different update stages and the more times of update operations on the same node, the longer the update waiting time delay. And the calculation complexity of the updating algorithm refers to the calculation complexity of the corresponding logic function in the process of switching the updating rule.

The results of comparing the embodiment of the present invention with the prior art are shown in table 2 below.

Table 2 comparison of the protocols herein with related protocols

1) And (4) applicability. Regardless of the complexity of the network topology, the classification timing sequence is updated according to the strict setting stage of the nodes on the new forwarding path and the old forwarding path. If the network configuration is simple, the updating efficiency is reduced by using the scheme. If a large number of common switches exist in configuration updating, different flow table operations need to be performed on the same switch at different updating stages, and the updating operation amount is increased. Therefore, the classification time sequence has certain application scene limitation and poor applicability. According to the scheme, on the basis of classification initialization, a loop search optimization model capable of reducing the number of updating rounds is designed, and the updating efficiency is guaranteed no matter whether the topological structure is complex or not.

2) And updating the time. The classification time sequence updating sets fixed stage updating, the node updating of the next stage can be carried out only after the node updating of the previous stage is finished, and the updating turn is fixed. If the update configuration is complex, the update time is also increased because the old packet is waited to flow out of the network and wait for the longest end-to-end delay. The node dependence of reverse update and bidirectional update is complex, so that the update round is multiple and the update time delay is high. The scheme combines a classification model and a loop search optimization model, the average updating round is close to the optimization updating, and the updating time is short.

3) And updating the operation complexity of the node. In the above embodiment, when the update configuration is complex, the classification timing update needs to perform different flow table update operations on a large number of nodes at different stages, and the number of update operations on a large number of nodes is large. Therefore, the operation complexity of the updating node of the scheme is high, and the time delay of updating waiting is long. Although the refined flow table entry operation enables the new and old data packets to be transmitted in parallel at a certain stage of updating, the load of the controller is reduced to a certain extent. But the time overhead is more important than the latency it generates. According to the scheme, only corresponding rule operation needs to be executed for the corresponding node set, the operation complexity is low, and the waiting time delay is short.

4) And (4) calculating complexity of an algorithm. It is assumed that the node information is stored using the adjacency matrix. In the optimization updating, initializing all node states, adding a new rule to detect whether a loop is introduced or not is O (n), and in the state conversion process, updating the node states and judging the loop of the next round of nodes are O (n). Thus, the time complexity of the optimization update is O (n)⁴). In the CSCU scheme, the total time complexity of classification initialization and judgment operation of nodes to be updated is O (n)²) Search for an updateable node as O (n), sinceThis total time complexity is O (n)³). Based on classification time sequence updating, only the matrix needs to be traversed for classification, then updating is carried out on different nodes, and the time complexity is O (n)²)。

In order to analyze and compare the updating effect of the related schemes under different conditions, the embodiment of the invention carries out simulation experiments based on the self-built SDN simulation platform, and the equipment used by the platform is a Ubuntu16.04 operating system, a RYU power controller and a mininet network simulator. The simulation experiment sets the number of experimental topological nodes, sets a source node and a destination node, simulates the change of a network topological structure by randomly changing the edge weight of adjacent nodes in each experiment, the edge weight of the adjacent nodes obeys (1-1000) uniform distribution, and acquires the path from the source node to the destination node through an OSPF algorithm after the weight is determined.

The relationship between the update round and the update time has been explained in detail in the above embodiment, and the update round of each experiment of the related research scheme is analyzed and compared by 300 simulation experiment tests, and the result is shown in fig. 5:

experimental results show that the required update round of the reverse update is generally more than other schemes, because the reverse update is only related to a new forwarding rule, and the update round is the new rule forwarding related node-1. Since the update phase is strictly designed for the classification timing update, the update round is fixed. The update rounds of the scheme (CSCU) of the embodiment of the invention are close to the optimal update, and are less than other schemes.

And step two, configuring and updating the cross degree. During the configuration updating process, the switches on the new and old paths account for the proportion of the total number of the switches related to the updating. The size of the configured update intersection level affects the update performance of the CSCU scheme. In the update process, the more switches are on the new path and not on the old path, the fewer switches are involved in the old rule and the more difficult it is to form a loop. Therefore, the fewer update rounds, the better the update performance. Instead, more rounds are required to complete the update. At this point, experimental analysis is carried out on the influence of the configured updating cross degree on the updating turn of the scheme. Results with crossover levels of 0% and 25% are not worth referencing since the likelihood of forming loops is low when the crossover level is low. Therefore, the results of the experiment are shown in FIG. 6, in which the degree of crossing was only 50%, 75%, and 100%.

For better comparison and analysis with classification timing sequence updating, the average updating turns of the scheme under the conditions of different cross degrees of configuration updating are obtained by integrating the experimental results of the CSCU scheme under the updating conditions of different cross degrees, and the experimental results show that the average updating turns of the scheme under the conditions of 0%, 50%, 75% and 100% of configuration updating cross degrees are less than the classification timing sequence updating, and the results are shown in FIG. 7.

And the node operation complexity in the updating process is directly embodied in the number of the switches with the operation times of the flow table items of the same switch being more than 1. The num _ large _1 is used to represent the number of switches with the operation times of the flow entry of the same switch being greater than 1, and obviously, the greater num _ large _1 is, the higher the operation complexity of the node in the updating process is. Aiming at the complexity of node operation, the embodiment of the invention designs a plurality of experiments, supposing that the number of nodes in each experiment is fixed to be 20 and the hop counts jump of the new path and the old path are the same, and performs test analysis on classification timing update to obtain the condition that the average value of num _ large _1 experiments changes along with the hop counts jump of the paths, wherein the experiment results are shown in fig. 8: (num _ large _1> is 2 because the new and old path source and destination nodes are set in the experiment).

Analysis experiments can show that num _ large _1 increases with increase of jump under the condition that the number of network nodes is constant. That is, as the hop count of the path increases, the number of nodes that are both nodes on the old and new paths increases. Therefore, the higher the node operation complexity of the update process. After the scheme is classified and initialized, the method is irrelevant to the complexity of updating configuration, and the complexity of updating node operation is low.

In summary, compared with the classification time sequence updating, the loop-free update Consistency Scheme (CSCU) based on the classification search has better scene applicability, lower node operation complexity and fewer update rounds, and can effectively reduce the update time delay and improve the update efficiency. In addition, in the aspect of updating round, the scheme has performance improvement in different aspects compared with other related research schemes.

According to the embodiment of the invention, through the demonstration of the updating example, the scheme and the related research scheme are compared and analyzed in the aspects of performances such as applicability, updating time and the like. Theoretical analysis is also carried out on the acyclic consistency updating of the scheme, and the feasibility of the scheme is proved. Meanwhile, the performance of the simulation experiment and the relevant scheme is compared and analyzed. The simulation experiment analyzes and compares the scheme and a related research scheme aiming at the updating turn, and also independently compares the updating cross degree and the node operation complexity of different configurations with a classification time sequence scheme. Simulation results show that the CSCU scheme has the advantages of high application scene applicability and low node updating operation complexity, and is a high-efficiency updating scheme capable of effectively reducing the updating required turns, updating time delay and reducing the calculation time complexity.

The second embodiment of the present invention relates to a system for updating loop-free consistency of an SDN flow table based on classified search, including:

a search module for searching the first set of nodes S_nodeAnd a fifth set of nodes S_updatedTo add to the fourth set of nodes T_iUpdating the fourth node set T_iUntil the first set of nodes S_nodeAnd a fifth set of nodes S_updatedThe difference set of (c) is null.

A third embodiment of the present invention relates to a server, as shown in fig. 9, including a processor (processor)810, a communication Interface (Communications Interface)820, a memory (memory)830 and a communication bus 840, where the processor 810, the communication Interface 820 and the memory 830 complete communication with each other through the communication bus 840. The processor 810 may invoke logic instructions in the memory 830 to perform the steps of the methods described in the embodiments above. Examples include:

where the memory and processor are connected by a communications bus, which may include any number of interconnected buses and bridges, connecting together the various circuits of the memory and one or more processors. The bus may also connect various other circuits such as peripherals, voltage regulators, power management circuits, and the like, which are well known in the art, and therefore, will not be described any further herein. A bus interface provides an interface between a communication bus and a transceiver. The transceiver may be one element or a plurality of elements, such as a plurality of receivers and transmitters, providing a means for communicating with various other apparatus over a transmission medium. The data processed by the processor is transmitted over a wireless medium via an antenna, which further receives the data and transmits the data to the processor.

The processor is responsible for managing the bus and general processing and may also provide various functions including timing, peripheral interfaces, voltage regulation, power management, and other control functions. And the memory may be used to store data used by the processor in performing operations.

A fourth embodiment of the present invention relates to a computer-readable storage medium storing a computer program. The computer program, when being executed by a processor, performs the steps of the method as described in the embodiments above.

That is, as can be understood by those skilled in the art, all or part of the steps in the method for implementing the embodiments described above may be implemented by a program instructing related hardware, where the program is stored in a storage medium and includes several instructions to enable a device (which may be a single chip, a chip, or the like) or a processor (processor) to execute all or part of the steps of the method described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.

It will be understood by those of ordinary skill in the art that the foregoing embodiments are specific examples for carrying out the invention, and that various changes in form and details may be made therein without departing from the spirit and scope of the invention in practice.

Claims

1. An SDN flow table loop-free consistency updating method based on classification search is characterized by comprising the following steps:

step S1, obtaining a first node set S to be updated in the network configuration_nodeSecond node set S without updating_unThird node set S only needing to delete old rules_{del_old}Fourth node set T that is updateable per round_iAnd an updated fifth set of nodes S_updated；

2. The SDN flow table loop-free consistency updating method based on classified search of claim 1, wherein the step S2 comprises:

3. The SDN flow table loop-free consistency updating method based on classified search as claimed in claim 1, wherein in step S3, the fourth node set T is updated_iAfter the node, further comprising:

4. The SDN flow table loop-free consistency updating method based on classified search as claimed in claim 1, wherein in step S3, the third node set S is updated_{del_old}And then, the method further comprises the following steps:

5. The SDN flow table loop-free consistency updating method based on classified search as claimed in claim 1, wherein in the step S4, a first set of nodes S is searched_nodeAnd a fifth set of nodes S_updatedTo add to the fourth set of nodes T_iThe method specifically comprises the following steps:

6. The SDN flow table loop-free consistency updating method based on classified search according to claim 1, wherein in the step S1:

7. An SDN flow table loop-free consistency updating system based on classified search is characterized by comprising:

8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program implements the steps of the SDN flow table loop-free consistency updating method based on classified search according to any one of claims 1 to 6.

9. A non-transitory computer readable storage medium, having stored thereon a computer program, which when executed by a processor, implements the steps of the SDN flow table loop-free consistency updating method based on classification search according to any one of claims 1 to 6.