Multiprocessor system fault diagnosis method and device, computer equipment and storage medium

Abstract

The multiprocessor system fault diagnosis method comprises the following steps that processors in a multiprocessor system serve as nodes, a communication link between two processors serves as an edge, a plurality of test objects are determined, and the test objects comprise a test node, two tested nodes and test links between the two test nodes and the tested nodes; the method comprises the following steps of: testing all test objects according to a test rule of a hybrid enhanced comparison model (HMM) * model to obtain symptoms; obtaining a mixed fault set according to symptoms obtained by testing, and screening the mixed fault set to obtain a mixed consistent fault set; and determining the faulted processor and / or communication link according to the mixed consistent fault set to obtain a diagnosis result. The invention provides a multiprocessor system fault diagnosis method and device, computer equipment and a storage medium, which can diagnose faults of processors and communication links in a multiprocessor system at the same time and are high in diagnosis degree.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-processor system fault diagnosis, and in particular to a multi-processor system fault diagnosis method, device, computer equipment and storage medium. Background Art

[0002] With the rapid development of high-performance computing technology and semiconductor technology, the number of processors integrated in supercomputer systems has continued to increase, and their structures have become increasingly complex. This trend has led to a sharp increase in the probability of processor failure. System-level fault diagnosis is a diagnostic technology that uses graph theory to provide fault diagnosis services. Traditional system-level fault diagnosis theories are mostly based on the situation where processor nodes fail, and a series of fault diagnosis strategies have been proposed accordingly, including one-step t-diagnosability, conditional t-diagnosability, g-good neighbor conditional diagnosability, g-extra conditional t-diagnosability, and R g Condition t-diagnosable, etc., these fault diagnosis strategies can only be used to diagnose faulty nodes.

[0003] In addition, the diagnostic process of system-level fault diagnosis needs to follow a certain fault diagnosis model. FPPreparatory et al. first proposed the classic PMC (Preparata, Metze, and Chien) model. In order to better evaluate the test results, Maeng et al. proposed a MM called the comparison model. * (Maeng and Malek) model. However, the PMC model is different from the MM * The models also only consider node failures, while ignoring the impact of communication link failures on the test. In practical applications, processor failures and communication link failures in multi-processor systems often coexist and occur simultaneously. Therefore, the existing technology that only considers node failures, that is, only considers processor failures, cannot accurately diagnose faults in multi-processor systems. Summary of the invention

[0004] In order to solve the deficiencies in the prior art, the present invention provides a multi-processor system fault diagnosis method, apparatus, computer equipment and storage medium, which can simultaneously diagnose faults of processors and communication links in the multi-processor system with high diagnostic accuracy.

[0005] In order to achieve the above object, the specific scheme adopted by the present invention is: a multi-processor system fault diagnosis method, comprising the following steps: The processors in the multi-processor system are taken as nodes, the communication links between two processors are taken as edges, and a plurality of test objects are determined, the test objects including a test node, two tested nodes, and test links between the two test nodes and the tested nodes; All test subjects were tested according to the test rules of the HMM* model (Hybrid Maeng and Malek, Hybrid Enhanced Comparative Model) to obtain symptoms; A mixed fault set is obtained according to the symptoms obtained from the test, and the mixed fault set is screened to obtain a mixed consistent fault set; The faulty processor and / or communication link is determined based on the hybrid consistent fault set to obtain a diagnosis result.

[0006] As a further optimization of the multi-processor system fault diagnosis method: the test rules of the HMM* model include: if the test node has no fault, the test process is reliable, and when the tested node and / or the test link has a fault, the symptom is that the test object has a fault, and when the tested node and the test link have no fault, the symptom is that the test object has no fault; If the test node fails, the test process is unreliable.

[0007] As a further optimization of the above multi-processor system fault diagnosis method: the specific method of testing all test objects according to the test rules of the HMM* model includes: The test node sends test information to the node under test through the test link; The node under test sends feedback information to the testing node according to the test information; The test node compares the two received feedback information and generates symptoms based on the comparison result.

[0008] As a further optimization of the above multi-processor system fault diagnosis method: a method for obtaining a mixed fault set according to the symptoms obtained by the test includes: Determine all possible results corresponding to the test object according to the test rules; Match the symptoms obtained from the test with the possible results. If the match is successful, add the symptoms to the mixed fault set corresponding to the test object; The mixed fault set is contained in the union of the node set and the edge set, where the node set consists of all nodes and the edge set consists of all edges.

[0009] As a further optimization of the above multi-processor system fault diagnosis method: the method of screening the mixed fault set to obtain the mixed consistent fault set includes: Take two mixed fault sets as screening targets. If the intersection of the two mixed fault sets is not an empty set, the two mixed fault sets are indistinguishable. Combine two indistinguishable mixed fault sets into a mixed consistent fault set; After traversing all mixed fault sets, multiple mixed consistent fault sets are obtained.

[0010] As a further optimization of the above multi-processor system fault diagnosis method: a method for determining a faulty processor and / or communication link according to a hybrid consistent fault set includes: Extracting nodes and / or edges that are faulty and unrelated in the mixed consistent fault set; The extracted processors corresponding to the nodes and / or the communication links corresponding to the edges are integrated into the diagnosis result.

[0011] A multi-processor system fault diagnosis device, used to implement the multi-processor system fault diagnosis method described above, the device comprising: A diagnosis management module is used to use processors in a multi-processor system as nodes, a communication link between two processors as an edge, and determine a plurality of test objects, the test objects including a test node, two tested nodes, and test links between the two test nodes and the tested nodes; Model management module, used to test all test objects according to the test rules of the HMM* model to obtain symptoms; A test management module is used to obtain a mixed fault set according to the symptoms obtained by the test, and to screen the mixed fault set to obtain a mixed consistent fault set, and The faulty processor and / or communication link is determined based on the hybrid consistent fault set to obtain a diagnosis result.

[0012] Computer equipment, including: Memory for storing computer programs; The processor is used to execute the computer program to implement the multi-processor system fault diagnosis method.

[0013] The computer-readable storage medium is used to store a computer program, and when the computer program is executed, the multi-processor system fault diagnosis method described above is implemented.

[0014] Beneficial effects:

[0015] The present invention combines three processors and two communication links into a test object based on the HMM* model, and can simultaneously test the processors and the communication links between the processors in the multi-processor system. In addition, the present invention is suitable for n-dimensional hypercube Q n The maximum number of mixed diagnosis faults is n-1, which can effectively improve the reliability and stability of the multi-processor system. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flow chart of a multi-processor system fault diagnosis method of the present invention; Figure 2 is an example diagram of two indistinguishable mixed fault sets; Figure 3It is a schematic diagram of the structure and expression of an n-dimensional hypercube; Figure 4 In the embodiment, the mixed diagnostic range is Schematic diagram when Figure 5 This is an example graph where nodes x and y are not adjacent and have no common neighbors when |V(F1ΔF2)|≥2; Figure 6 It is an illustration of the edge value of node y when edge (x,y)∈R. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] See also Figure 1 , a multi-processor system fault diagnosis method, including S1 to S4.

[0019] S1. The processors in the multi-processor system are regarded as nodes, the communication links between two processors are regarded as edges, and multiple test objects are determined. The test objects include a test node, two tested nodes, and test links between the two test nodes and the tested nodes.

[0020] More specifically, a multiprocessor system can be represented by an undirected graph G(V,E), where V(G) and E(G) represent the node set and edge set of the system respectively. For any node x∈V(G), the node set and edge set adjacent to node x are represented by N V (x) and N E (x). The degree of node x refers to the number of nodes adjacent to x, denoted by d v (x) = |N(x)|. The degree of a system G refers to the minimum degree of all nodes in the system, denoted by δ(G). Obviously, for any node x∈V(G), there exists d(x)≥δ(G). In addition, for a mixed set containing nodes and edges V(F) refers to the node set of the mixed set F, and E(F) refers to the edge set of the mixed set F. Therefore, for a mixed set F, |F| = |V(F)| + |E(F)|.

[0021] Given an arbitrary mixture set If GF is disconnected or a trivial graph, then F is called a vertex-edge mixed cut set of system G. Let u,v∈V(G), and the set of common neighbors of u and v in G is denoted by cn(u,v). The maximum number of common neighbors of system G is the maximum number of neighbors between any two nodes in G, denoted by

[0022] A test object can be represented as t(b,c;a), where node a is called a test node, nodes b and c are called tested nodes, edges (a,b) and (a,c) are called test edges, and nodes a,b,c∈V(G), b and c are two different neighboring nodes of node a. In addition, it is assumed that all edges connected to the faulty nodes are fault-free, and the test results are reliable when the test nodes are fault-free. The test results of the test object are represented by 0 or 1. For example, if the test results of node a testing nodes b and c are correct, then the test result is 0, otherwise the test result is 1.

[0023] S2. All test objects are tested according to the test rules of the HMM* model (Hybrid Maeng and Malek, Hybrid Enhanced Comparison Model) to obtain symptoms. The test rules of the HMM* model include: if the test node is fault-free, the test process is reliable, and when the tested node and / or the test link are faulty, the symptom is that the test object is faulty, and when the tested node and the test link are not faulty, the symptom is that the test object is fault-free; if the test node is faulty, the test process is unreliable. More specifically, the test rules are shown in Table 1.

[0024] Table 1 HMM* model test rules table

[0025] In Table 1, the test result of the test object t(b,c;a)=1 indicates that at least one of the nodes a,b,c and the test edges (a,b), (a,c) is faulty; t(b,c;a)=0 indicates that all nodes and edges are faulty or the test node a is faulty.

[0026] The specific method of testing all test objects according to the test rules of the HMM* model includes S21 to S23.

[0027] S21. The testing node sends testing information to the node under test through the testing link.

[0028] S22. The node under test sends feedback information to the testing node according to the test information.

[0029] S23. The test node compares the two pieces of feedback information received and generates a symptom according to the comparison result.

[0030] For a mixed set If there are faults in the test nodes, tested nodes and / or test edges, the mixed set can be converted into a mixed fault set and still be Since the test results sent by the faulty nodes are unreliable, the mixed fault set F may produce symptoms. The set of all symptoms that can be produced by F can be represented by σ(F).

[0031] S3, obtaining a mixed fault set according to the symptoms obtained by the test, and screening the mixed fault set to obtain a mixed consistent fault set. The method of obtaining a mixed fault set according to the symptoms obtained by the test includes S31 to S33.

[0032] S31. Determine all possible results corresponding to the test object according to the test rules.

[0033] S32, matching the symptoms obtained from the test with the possible results, and if the match is successful, adding the symptoms to the mixed fault set corresponding to the test object.

[0034] S33. The mixed fault set is contained in the union of the node set and the edge set. The node set consists of all nodes, and the edge set consists of all edges.

[0035] The method for screening mixed fault sets to obtain mixed consistent fault sets includes: taking two mixed fault sets as screening targets, if the intersection of the two mixed fault sets is not an empty set, then the two mixed fault sets are indistinguishable; combining the two indistinguishable mixed fault sets into a mixed consistent fault set; and traversing all mixed fault sets to obtain multiple mixed consistent fault sets. More specifically, for two arbitrarily different mixed fault sets in a multiprocessor system G, If and only if When , (F1, F2) is indistinguishable, otherwise (F1, F2) is distinguishable. For example, given a multiprocessor system consisting of 5 nodes and 4 edges, such as Figure 2 As shown, let F1 = {(a,b),c}, F2 = {b,c}. Obviously, both F1 and F2 can generate the syndrome set of σ(b,c;a) = σ(b,e;a) = σ(b,d;a) = σ(c,e;a) = σ(c,d;a) = 1, σ(d,e;a) = 0, so That is to say, F1 and F2 are indistinguishable. At the same time, since the fault nodes in the fault set F1 and F2 are not associated with the red fault edges, (F1, F2) can be called a mixed consistent fault set.

[0036] Furthermore, if are any two different mixed-consistent fault sets in a multiprocessor system G, and |F1|,|F2|≤t h, if (F1, F2) are distinguishable, then the multiprocessor system G is a mixed t h -Diagnosable system. Based on HMM * Model, the mixed diagnostic degree of system G is denoted as Denotes a multiprocessor system G as a mixed t h - Diagnostic system time h The maximum value that can be obtained.

[0037] S4. Determine the faulty processor and / or communication link according to the hybrid consistent fault set to obtain a diagnosis result. Because the tested nodes and test edges of the fault in the hybrid consistent fault set are independent, the faulty nodes and / or edges can be determined from the hybrid consistent fault set, that is, the faulty processor and / or communication link can be determined. The method for determining the faulty processor and / or communication link according to the hybrid consistent fault set includes S41 to S42.

[0038] S41. Extracting nodes and / or edges that are faulty and unrelated in the mixed consistent fault set.

[0039] S42: Integrate the extracted processors corresponding to the nodes and / or the communication links corresponding to the edges into a diagnosis result.

[0040] In a mixed-consistent fault, each element is a node or an edge, that is, the mixed-consistent fault set is composed of processors and communication links. The failed processor and communication link can be determined by extracting the identifiers of the processor and communication link in the system.

[0041] In addition, a multiprocessor system can be represented as a hypercube network. An n-dimensional hypercube network Q n Total 2 n nodes (x1x2…x n ), for any x i ∈{0,1}, i∈[1,n] and n·2 n-1 For example, a 4-dimensional hypercube Q4 can be represented by a 4-dimensional binary string, such as Figure 3 As shown. At the same time, for Q n Any node x∈V(Q n ), both exist |N V (x)|=d v (x) = δ(Q n )=n. In addition, let u,v∈V(Q n ), if u and v have common neighbors, then the number of their common neighbors cn(u,v)=2, otherwise cn(u,v)=0.

[0042] In multiprocessor system fault diagnosis, the diagnostic degree is a very important parameter.n , based on the diagnostic method of the present invention, its mixed diagnostic degree is for n≥3. Compared with the prior art, the hybrid diagnosis of the present invention is higher and can better perform fault diagnosis on a multi-processor system.

[0043] More specifically, the method for determining the degree of diagnosis of the present invention is as follows.

[0044] First, let the node x∈V(Q n ),δ(x)=n,N V (x)={y1,y2,……,y n}. Let F1={x}, F2={(x,y1),...,(x,y n )},like Figure 4 As shown. It can be seen that |F1|=1, |F2|=n. Obviously, (F1, F2) is a mixed consistent fault pair, F1 and F2 in HMM * Therefore, the system G is not a mixed t h - Diagnosable system. Therefore, for n≥3.

[0045] Second, suppose there are two different sets of mixed consistent faults And in HMM * The model is indistinguishable, where |F1|,|F2|≤n-1. Let R=V(Q n )∪E(Q n )-F1-F2, because |V(F1)|≤|F1|≤n-1, |V(F2)|≤|F2|≤n-1, so when n≥3, we have Right now And because F1≠F2, or Depending on whether there is a node in F1ΔF2, the following two cases are discussed.

[0046] Case A, there is a node in F1ΔF2, due to Depending on whether V(F1ΔF2) has neighbors in R, it is further divided into the following two sub-cases.

[0047] Case A1, when When , according to the number of nodes in F1ΔF2, it is divided into two more specific cases: |V(F1ΔF2)|=1 and |V(F1ΔF2)|≥2.

[0048] Case A11, let |V(F1-F2)|=1, Then V(F1∩F2)=F2. Let the node x∈V(F1-F2). and Therefore N V (x)∈F1∩F2. Then |F2|=|N V (x)|≥n, which contradicts the assumption that |F2|≤n-1.

[0049] Case A12, let x∈V(F1-F2), y∈V(F2-F1). Because so and Therefore N V (x)∈F1∪F2,N V (y)∈F1∪F2, such as Figure 5 As shown. And because x,y∈V(Q n ), so when n ≥ 3, cn(x, y) ≠ 0 or cn(x, y) = 2. Therefore, regardless of whether nodes x and y are adjacent or have common neighbors, |V(F1) ∪V(F2)| = |N V (x)∪N V (y)|≥2n, which contradicts the assumption that |F1|+|F2|≤2n-2. Similarly, when |V(F1-F2)|≥2 or |V(F2-F1)|≥2, a similar contradiction can be obtained: |V(F1)∪V(F2)|≥2n.

[0050] Case A2, let node x∈V(F1-F2), and let any node y∈N V(R) (x) and edge (x,y)∈R. Since F1 and F2 are mixed consistent fault sets, N V(R) The connecting edge of (x) belongs to R∪F2-F1, The connecting edges belong to R∪F2-F1, The connected edges of belong to R. And The connecting edge belongs to F2-F1, The connecting edges belong to R∪(F1-F2), The connecting edge belongs to R. And because F1 and F2 are indistinguishable, N V(R) The connecting edges of (y) can only belong to F2, and N V(R) (y) The edge connecting F2-F1 and The total number of connected edges belonging to R is at most 1. That is, Figure 6 The situation within the blue dashed line occurs at most once, otherwise F1 and F2 can be distinguished, so there is By N V (x)-y distribution to make judgments one by one.

[0051] Case A21, if there exists any point z∈N V(R)(x), when edge (x,z)∈R, then because F1 and F2 are indistinguishable mixed consistent fault sets, the remaining neighbors of node z and the connecting edges are Figure 6 The neighboring nodes of the middle node y are similar to the connecting edges, and also have Since nodes y, z have a common neighbor x and y, z∈V(Q n ), then cn(y,z)=2. Therefore, when n≥3, This contradicts the assumption that |F1|+|F2|≤2n-2.

[0052] Therefore That is N V(R) The edges connecting (x)-y will all belong to F2-F1. And because The connecting edges belong to R∪F2-F1, The connecting edges of belong to R. Then there exists Therefore, when n ≥ 3, This contradicts the assumption that |F1|+|F2|≤2n-2.

[0053] Case A22, because That is N V (x)-y∈F1∪F2, we have Therefore, when n ≥ 3, This also contradicts the assumption that |F1|+|F2|≤2n-2. In this case, let any node x∈V(F1-F2). Because F1, F2 are indistinguishable mixed consistent fault sets and N V(R) The connecting edge of (V(F1ΔF2)) does not belong to R, so N V(R) The connecting edges of (x) all belong to F2-F1, The connected edges belong to R, The connecting edges belong to R∪(F2-F1). It can be found that If there is any point So that the connecting edge (x, y) ∈ R. Then due to N V(R) The connecting edge of (V(F1ΔF2)) does not belong to R, so the node y is similar to x, that is, N V(R) The connecting edges of (y) all belong to F2-F1, The connected edges belong to R, The connecting edges belong to R∪(F2-F1). Similarly, there exists

[0054] Therefore, when n ≥ 3, This contradicts the assumption that |F1|+|F2|≤2n-2. The connecting edges can only belong to F2-F1, This contradicts the assumption that |F2|≤n-1.

[0055] Therefore, there is Then N V (x)∈R∪F2. And N V(R) The connecting edge of (x) belongs to F2-F1, The connected edges belong to R. This also contradicts the assumption that |F2|≤n-1.

[0056] Case B, there is no node in F1ΔF2. or So In this case, the following two situations are discussed.

[0057] Case B1, if |E(F1ΔF2)|=1, let any edge (u,v)∈E(F1-F2), Because F1 is a mixed consistent fault set, Also because Therefore u,v∈R and there exists Taking node u as an example, since F1 and F2 are indistinguishable and so The connecting edges of can only belong to R; N V(R) The connecting edge of (u) can only belong to F1∩F2, so there is This contradicts the assumption that |F1|≤n-1.

[0058] Case B2, if |E(F1ΔF2)|≥2, assume that there is an edge in F1-F2 and F2-F1 and these two edges must have a common endpoint, otherwise it is similar to Case 2.1. Let any edge (u,v)∈E(F1-F2), (u,w)∈E(F2-F1) and u,v,w∈R. It is not difficult to find that although nodes v and w have a common neighbor u, because v,w∈V(Q n ), so cn(v,w)=2. Because F1, F2 are mixed consistent fault sets and So N V(R) (u), N V(R) (v) with N V(R) The connecting edges of (w) belong to R∪F1∪F2; and The connecting edge belongs to R. But because F1 and F2 are indistinguishable and there are already edges (u,v)∈E(F1-F2) and (u,w)∈E(F2-F1), so N V(R) The connecting edges of (u) can only belong to F1∩F2; N V(R) (v) The connecting edge belongs to F2 and N V(R) The connecting edge of (w) belongs to F1 and Otherwise F1 and F2 are distinguishable. Therefore, when n ≥ 3, there exists At this time, When n ≥ 3, it contradicts the assumption that |F1|+|F2|≤2n-2.

[0059] In summary, the assumption is not true, that is, for n≥3.

[0060] Finally, combine for n ≥ 3 and For n≥3, the diagnostic degree of the present invention can be determined as for n≥3. That is to say, for a multi-processor system including n processors, the maximum number of faulty processors and communication links that can be diagnosed by the system using the present invention is n-1. When one or some processors fail, the system can quickly locate the faulty processors, and then adopt corresponding means to avoid the instability caused by the faulty processors, so that the stability and reliability of the system are fully improved.

[0061] The present invention further provides a multi-processor system fault diagnosis device for implementing the multi-processor system fault diagnosis method described above. The device includes a diagnosis management module, a model management module and a test management module.

[0062] The diagnostic management module is used to use the processors in the multi-processor system as nodes, the communication links between two processors as edges, and determine multiple test objects, the test objects including a test node, two tested nodes and test links between the two test nodes and the tested nodes.

[0063] The model management module is used to test all test objects according to the test rules of the HMM* model to obtain symptoms.

[0064] The test management module is used to obtain a mixed fault set according to the symptoms obtained from the test, screen the mixed fault set to obtain a mixed consistent fault set, and determine the faulty processor and / or communication link according to the mixed consistent fault set to obtain a diagnosis result.

[0065] The present invention also provides a computer device, including a memory and a processor.

[0066] Memory is used to store computer programs. The memory, as a carrier for storing resources, can be a read-only memory, random access memory, a disk or an optical disk, etc. The resources stored thereon can include an operating system, a computer program, etc., and the storage method can be temporary storage or permanent storage. The operating system is used to manage and control various hardware devices and computer programs on the electronic device, which can be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program that can be used to complete the multi-processor system fault diagnosis method performed by the electronic device disclosed in any of the aforementioned embodiments, the computer program can further include a computer program that can be used to complete other specific tasks.

[0067] The processor is used to execute the computer program to implement the multi-processor system fault diagnosis method. The processor can be a general-purpose processor product based on X86, IA64, RISC, MIPS, ARM and other architectures.

[0068] Finally, the present invention also provides a computer-readable storage medium for storing a computer program, and the computer program implements the multi-processor system fault diagnosis method when executed.

[0069] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A multi-processor system fault diagnosis method, characterized in that: The steps include: The processors in the multi-processor system are taken as nodes, the communication links between two processors are taken as edges, and a plurality of test objects are determined, the test objects including a test node, two tested nodes, and test links between the two test nodes and the tested nodes; All test subjects were tested according to the test rules of the HMM* model (Hybrid Maeng and Malek, Hybrid Enhanced Comparative Model) to obtain symptoms; A mixed fault set is obtained according to the symptoms obtained from the test, and the mixed fault set is screened to obtain a mixed consistent fault set; The faulty processor and / or communication link is determined based on the hybrid consistent fault set to obtain a diagnosis result.

2. The multiprocessor system fault diagnosis method according to claim 1, characterized in that: The test rules for the HMM* model include: If the tested node has no faults, the test process is reliable, and when the tested node and / or the tested link has faults, the symptom is that the tested object has faults, and when neither the tested node nor the tested link has faults, the symptom is that the tested object has no faults; If the test node fails, the test process is unreliable.

3. The multiprocessor system fault diagnosis method according to claim 1, characterized in that: The specific method of testing all test objects according to the test rules of the HMM* model includes: The test node sends test information to the node under test through the test link; The node under test sends feedback information to the testing node according to the test information; The test node compares the two received feedback information and generates symptoms based on the comparison result.

4. The multiprocessor system fault diagnosis method according to claim 1, characterized in that: Methods for obtaining mixed fault sets based on symptoms obtained from tests include: Determine all possible results corresponding to the test object according to the test rules; Match the symptoms obtained from the test with the possible results. If the match is successful, add the symptoms to the mixed fault set corresponding to the test object; The mixed fault set is contained in the union of the node set and the edge set, where the node set consists of all nodes and the edge set consists of all edges.

5. The multiprocessor system fault diagnosis method according to claim 1, characterized in that: Methods for screening mixed fault sets to obtain mixed consistent fault sets include: Take two mixed fault sets as screening targets. If the intersection of the two mixed fault sets is not an empty set, the two mixed fault sets are indistinguishable. Combine two indistinguishable mixed fault sets into a mixed consistent fault set; After traversing all mixed fault sets, multiple mixed consistent fault sets are obtained.

6. The multiprocessor system fault diagnosis method according to claim 1, characterized in that: A method for determining a failed processor and / or communication link based on a hybrid consistent fault set includes: Extracting nodes and / or edges that are faulty and unrelated in the mixed consistent fault set; The extracted processors corresponding to the nodes and / or the communication links corresponding to the edges are integrated into the diagnosis result.

7. A multi-processor system fault diagnosis device, used to implement the multi-processor system fault diagnosis method according to any one of claims 1 to 6, characterized in that: The device comprises: A diagnosis management module is used to use processors in a multi-processor system as nodes, a communication link between two processors as an edge, and determine a plurality of test objects, the test objects including a test node, two tested nodes, and test links between the two test nodes and the tested nodes; Model management module, used to test all test objects according to the test rules of the HMM* model to obtain symptoms; A test management module is used to obtain a mixed fault set according to the symptoms obtained by the test, and to screen the mixed fault set to obtain a mixed consistent fault set, and The faulty processor and / or communication link is determined based on the hybrid consistent fault set to obtain a diagnosis result.

8. Computer device, characterized in that include: Memory for storing computer programs; A processor, configured to execute the computer program to implement the multi-processor system fault diagnosis method as claimed in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that Used to store a computer program, which, when executed, implements the multi-processor system fault diagnosis method as described in any one of claims 1 to 6.

Images