JP6632558B2 - Network verification optimization method, program and apparatus - Google Patents

Description

  The present invention relates to a network verification optimization method, a program, and an apparatus, and more particularly, to a network verification optimization method, a program, and an apparatus capable of maximizing verification coverage within a limited number of verifications by reducing the number of verification steps.

  Conventionally, various techniques for automating a fault test have been proposed for improving the efficiency of a network fault test. Non-Patent Literature 1 describes a method for verifying the behavior and operation of a network in a verification environment simulating a commercial network before performing expansion, modification, or upgrading of a network OS on an existing commercial network. There has been proposed a system for automatically constructing a verification environment, performing verification, and analyzing results.

"A Proposal on Verification Automation System Design for Networks Utilizing White Box Switches and OSS", Emi Shibuya, Hidehiko Kawakami, Teruyuki Hasegawa, Hirozumi Yamaguchi, IEICE Technical Report, vol. 116, no. 111, NS2016-34, pp. 35-40, June 2016.

  Non-Patent Literature 1 greatly reduces the number of steps of verification work conventionally performed manually by a verification worker by automatically performing verification, implementation of verification, analysis of results, and the like. However, since it can be executed automatically, verifiers will aim to improve verification performance by comprehensively verifying the entire network that could not be handled manually before, so verification items will increase further It is possible to do.

  Furthermore, since the verification time is limited, there is a problem that an exhaustive failure test cannot always be completed within a given verification time.

  An object of the present invention is to solve the above technical problems and to provide a network verification optimization method, a program, and an apparatus that can maximize verification coverage within a limited number of verifications and enable efficient verification. It is in.

  In order to achieve the above object, the present invention is characterized in having the following configuration.

  (1) The network fault test apparatus of the present invention provides a faulty unit (ci, cj, ek) related to a set of a client pair (ci, cj) connected to the network and one link ek on the route. Means for calculating a similarity index value when the link ek of the failed unit is a failure location, and similarity for evaluating the similarity between the failed units (ci, cj, ek) based on the similarity index value If the evaluation means and verification of mutually similar failure units (ci, cj, ek) are performed on any one of the similar failure units (ci, cj, ek), other similar failure units The faulty unit is considered to have been executed, and is provided with a unit for extracting a fault location that maximizes the sum of the number of verifications performed and the number of verifications considered.

  (2) The network failure test method according to the present invention provides a method for each failure unit (ci, cj, ek) relating to a set of a client pair (ci, cj) connected to the network and each link ek on the route. A first procedure for calculating a similarity index value when the link ek of the failed unit is a failed location, and for each failed location on the network, the similarity between the failed units including the failed location in the path is referred to as the similarity. A second procedure for evaluating based on the degree index value, a third procedure for deriving a set D of failure units similar to each other for each failure location, and a set D of failure units for the failure location to be verified are aggregated into a set Z. A fourth procedure for setting a constraint condition on the number of verifications, and a sixth procedure for calculating a fault location for maximizing the number of faulty units of the set Z as an integer programming problem using the second to fifth procedures. And provided.

  (3) The similarity index value includes at least one of the path length of the route before the occurrence of the failure, the path length of the route re-searched after the occurrence of the failure, and the network hierarchy at the location of the failure.

  (4) The similarity index value includes at least one of a link band, a switch connection order, a delay, a link number, an order, and a link importance.

  According to the present invention, the following effects are achieved.

  (1) A failure unit whose verification is to be substantially duplicated is searched for based on the similarity of its topology, and in a failure unit set having a similar topology, if one failure unit is verified, another verification unit is verified. Is also considered to have been verified, duplicate verification is avoided, the number of verifications is reduced, and efficient verification becomes possible.

  (2) Within the pre-specified number of verifications, the fault location that maximizes the sum of the “number of verifications performed” and the “number of deemed verifications” is searched, so that the verification coverage is maximized within the limited number of verifications And efficient verification becomes possible.

  (3) Since the similarity of the topology depends on the index value to be adopted, an appropriate similarity judgment can be made for each network by changing the similarity index value according to the network to be verified.

It is a functional block diagram of a network verification optimization device concerning one embodiment of the present invention. FIG. 7 is a diagram illustrating a state where a route changes before and after a link failure occurs. FIG. 3 is a diagram for explaining similarity of network topologies. FIG. 14 is a diagram illustrating an example of deriving a similar set set D (e).

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. Here, the outline of the present invention will be described first, and then its embodiments will be described.

  In the present invention, a commercial network operated by a network provider is to be verified, and a topology (sub-topology) having a similar shape, hierarchy, and the like exists on such a network. The topology is composed of a plurality of layers k as shown in FIG. 2, and a client c is accommodated in a leaf of the topology.

  Generally, in the single link failure test, verification is attempted by causing a failure once for each link on the network. The communication verification between the client pairs is performed for each of the failures of the links constituting the path for the path between all the client pairs. Therefore, the minimum unit of verification is a set of a client pair and a failed link on the path before the failure. In the present embodiment, such a set is expressed as a “failure unit”, and is adopted as one index for quantitatively evaluating the similarity of the minimum unit topology, as described later in detail.

  When the path length of the route between the client pair ci and cj before the occurrence of the failure is Hij, as shown in the following equation (1), verification is required F times in order to comprehensively verify the network.

  On the other hand, in the present invention, in order to obtain the maximum verification coverage (coverage) with a limited number of verifications, if the verification is performed only once for the failure units that are similar to each other, the other failure units can be verified. Based on the idea that can be regarded as verified, it is calculated which link can be subjected to a failure test to obtain the maximum verification coverage with the minimum number of verifications. That is, the present invention considers deriving a combination of links (failure points) for performing a failure test.

  In the present invention, the concept of the failure unit is introduced as an index for quantitatively evaluating the similarity of the topology for each minimum unit of the network verification. Then, the topology in which the failed units are similar is evaluated as similar from the viewpoint of verification, and for the failed units that are similar to each other, it is considered that if one of the failed units is verified, the other failed units are also verified. A failure location that maximizes the sum of the “verification count” and the “verification count considered to be performed (deemed verification count)” is derived.

  FIG. 1 is a functional block diagram showing a configuration of a main part of a verification optimization system according to an embodiment of the present invention. Here, components unnecessary for the description of the present invention are omitted. Such a system can be configured by mounting an application (program) for realizing each function described later on a general-purpose computer or server. Alternatively, a part of the application may be configured as a dedicated device or a single-purpose device in which hardware or ROM is used.

  In the verification optimization system 1, the route search unit 10 calculates the shortest route for each combination of clients (client pair ci, cj) connected to the network based on a separately provided network topology.

  The failed unit registration unit 20 registers the failed unit (ci, cj, e) which assumes that the client pair (ci, cj) and each link on the route are the failed link e. Therefore, for a client pair (ci, cj) with a path length of n, n failure units (ci, cj, e1), (ci, cj, e2) ... (ci, cj, en) must be registered become.

  The similarity index value calculation unit 30 calculates the similarity index value of the network topology when a failure occurs in the link e for all the failed units. The commercial network to be verified in this embodiment is generally of a layered structure as shown in FIG. 2, and it is known that the same behavior and almost the same device are used for each layer. . Therefore, if the same hierarchy and the same path length before and after the occurrence of the failure are the same, it can be determined that the path passes through a path having almost the same behavior. Therefore, in the present embodiment, the following three are calculated as similarity index values. You.

q: Path length of route before failure
r: Path length after a failure occurs on link e on the route
k: Network hierarchy of the failed link e on the route

  Here, the path length q is the path length of the shortest path calculated for each client pair (ci, cj). The path length r is a bypass route between the client pair (ci, cj) recalculated in consideration of the failure link e.

  The similarity evaluation unit 40 compares the three similarity index values q, r, and k calculated for each failure unit (ci, cj, e) between all failure unit pairs. Then, the failed units having the same three similarity index values q, r, and k are evaluated as similar (similarity = 1), and if at least one unit is different, dissimilar (similarity = 0) is evaluated. Units are aggregated into the same failed unit set.

  The optimization unit 50 determines that the failure unit (ci, cj, e) aggregated in the same failure unit set is verified only for one of the failure units and verified for the other failure units. Assuming, a set of failure occurrence points (links) that maximizes the verification coverage within the limited number of times of verification, that is, the sum of “performed verification times” and “deemed verification times”, is derived. In order to complete the test within a limited verification time, the number of verifications N can be specified.

  The derived set of fault occurrence locations is notified to the verification enforcement unit 2, and a fault is automatically or manually generated at each fault occurrence location to confirm communication between client pairs.

  Next, let us consider a formulation as an optimization problem in which the optimization unit 50 computes a set of fault occurrence locations by a computer as an optimal solution.

  Here, it is assumed that a network of layer k is configured by S switches V = {v1, v2... VS} and L links E = {e1, e2,. Such a network can be expressed as E⊆V × V in a network graph G = (V, E) composed of the link E and the switch V.

  Assuming that the number of clients is M, all clients belong to the client set C = {c1, c2,... CM} and are accommodated under each switch. A set of links e∈E belonging to the shortest path between the client pair ci and cj is represented by Primary (ci, cj), and this is defined as a path before a failure has occurred. Then, a link set belonging to a detour path recalculated due to a failure of the single link e (e∈Primary (ci, cj)) belonging to Primary (ci, cj) is defined as Secondary (ci, cj, e).

  For example, the link set Primary (c1, c3) = {17, 11, 5, 1, 7, 15, 19} of the pre-failure path between the client pair c1 and c3 in FIG. When a failure occurs, the link set of the detour path is Secondary (c1, c3, e5) = {17, 13, 6, 4, 8, 16, 19}. The pre-failure route and the detour route can both be calculated using the shortest route calculation algorithm Dijkstra or the like.

  Here, for a failure of a certain link e (failure link e), a failure unit (cx, cy, e) relating to an arbitrary client pair cx, cy including the failure link e in the path before the failure, and an arbitrary client The similarity Sim ((cx, cy, e), (ca, cb, ez)) with the faulty unit (ca, cb, ez) for the pair ca, cb and any faulty link ez is calculated by the above-mentioned 3 of each faulty unit. If all the similarity index values q, r, and k match, they are evaluated as similar (= 1), and if even one does not match, they are evaluated as dissimilar (= 0). Therefore, the similarity Sim between the two failure units (cx, cy, e) and (ca, cb, ez) can be expressed by the following equation (2).

  For example, the two faulty units (c1, c3, e11) and (c2, c3, e12) in FIG. 2 are similar because q = 7, r = 7, and k = 3, and Sim ((1, 3, 11, 11) ), (2, 3, 12)) = 1. In the present embodiment, the similarity Sim of the above equation (2) is calculated between all the failed units.

  Based on the similarity calculation result, a similar set set D (e) for the link e failure is derived. D (e) can be represented by the following equation (3).

  As an example, FIG. 4 shows the derivation result of D (e) in FIG. In FIG. 4, two failure units (1, 3, 1) and (2, 3, 1) are similar to each other for the link e1 failure, and four failure units (1, 3, 5) for the link e5 failure. , (2,3,5), (1,3,7) and (2,3,7) are similar to each other.

  Furthermore, a binary function down (e) that indicates whether or not to perform the verification of the link e is defined. If "1" is defined as performing the verification and "0" is defined as not performing the verification, the verification test of the link e is performed. A set Z (e) of trial similarity sets to be implemented can be expressed by the following equation (4).

  Here, the constraint condition that the number of times of verification is N or less can be expressed by the following equation (5).

  Under the constraints of the above equations (2) to (5), the following equation (6) is calculated as the objective function for the combination that maximizes the total number of similar sets.

  Since the above equations (2)-(6) are expressed as integer computation problems, it is possible to calculate the optimal solution with a general-purpose solver (software such as IBM ILOG CPLEX).

  In this formula, the similarity based on the path length before and after the occurrence of the failure and the hierarchy of the failure link e has been described, but the present invention classifies and defines the similarity with different indices, and provides other It is possible to apply the similarity method of.

  In the above embodiment, the topology similarity index value includes the path length q of the pre-harm occurrence path, the path length r of the link e on the path after a failure has occurred, and the network of the failure link e on the path. Although it has been described that the layer k is adopted, the present invention is not limited to this, and instead of the similarity index value or in addition to the similarity index value, a link band, a switch connection The order, delay, link importance, etc. may be adopted.

  DESCRIPTION OF SYMBOLS 1 ... Verification optimization system, 2 ... Verification enforcement part, 10 ... Route search part, 20 ... Failure unit registration part, 30 ... Similarity index value calculation part, 40 ... Similarity evaluation part, 50 ... Optimization part

Claims (7)

For each failure unit (ci, cj, ek) related to a set of a client pair (ci, cj) connected to the network and each one link ek on the route, a case where the one link ek is a failure location Means for calculating a similarity index value;
A similarity evaluation means for evaluating the similarity between the failure units (ci, cj, ek) based on the similarity index value;
If the verification of mutually similar failure units (ci, cj, ek) is performed on any one of the similar failure units (ci, cj, ek), other similar failure units are performed. A means for extracting a failure location that maximizes the sum of the number of verifications performed and the number of verifications deemed to have been performed, the verification optimization apparatus for a network.   2. The method according to claim 1, wherein the similarity index value includes at least one of a path length of a route before the occurrence of the failure, a path length of a detour route searched again after the occurrence of the failure, and a network hierarchy of the failure location. A network verification optimization device as described.   3. The network verification optimization apparatus according to claim 1, wherein the similarity index value includes at least one of a link band, a switch connection order, a delay, and a link importance. For each failure unit (ci, cj, ek) related to a set of a client pair (ci, cj) connected to the network and each one link ek on the route, a case where the one link ek is a failure location A first step of calculating a similarity index value,
A second procedure for evaluating, for each fault location on the network, the similarity between the faulty units including the fault location on the path based on the similarity index value;
A third procedure for deriving a set D of failure units similar to each other for each failure location;
A fourth procedure of aggregating the set D of the faulty units relating to the fault location to be verified into the set Z;
A fifth procedure for setting constraints on the number of verifications,
A sixth procedure of calculating a fault location that maximizes the number of faulty units of the set Z as an integer programming problem using the second to fifth procedures.   5. The method according to claim 4, wherein the similarity index value includes at least one of a path length of the route before the occurrence of the failure, a path length of the detour route searched again after the occurrence of the failure, and a network hierarchy of the failure location. Verification optimization method for the described network.   6. The network verification optimization method according to claim 4, wherein the similarity index value includes at least one of a link band, a switch connection order, a delay, and a link importance.   A network verification optimization program for causing a computer to execute the network verification optimization method according to any one of claims 4 to 6.