JP5365273B2 - Information processing system, monitoring method, and monitoring program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an information processing system, a monitoring method and a monitoring program for monitoring whether or not a system is in normal operation. <P>SOLUTION: An information processing system is provided with: a first CPU 80 for executing a plurality of processes 70, 71 and 72; a second CPU 81 for monitoring the plurality of processes 70, 71 and 72 executed by the first CPU 80; an analyzing part 60 operated by the second CPU 81 for extracting the mutual dependence of the plurality of processes 70, 71, and 72 based on executing situations of the plurality of processes 70, 71 and 72; and a monitoring part 61 operating in the secure CPU 81, and detecting abnormalities of the operations of the plurality of processes 70, 71 and 72 based on dependence extracted by the analyzing part 60. <P>COPYRIGHT: (C)2010,JPO&amp;INPIT

Description

  The present invention relates to an information processing system, a monitoring method, and a monitoring program.

  Various information processing systems are required to operate normally. In particular, an in-vehicle embedded system is required to have high reliability, high robustness, high operation predictability, and the like because a malfunction is related to human life. In order to satisfy such a requirement, it is necessary to monitor whether the system is operating normally.

Patent Document 1 detects a failure currently occurring in the information processing system by comparing the correlation of the monitoring data when the information processing system is operating normally with the latest monitoring data, Disclosed is a technology for predicting a failure that may occur in the future in an information processing system from the similarity between the correlation of monitoring data when an abnormality occurs in the information processing system and the correlation of recently obtained monitoring data ing.
However, the technique disclosed in Patent Document 1 is not related to a multi-CPU (Central Processing Unit) system.

  Patent Document 2 also discloses information collected when the first process actually calls the second process, and information collected when the second process is called from the first process and starts processing. By analyzing the linkage status of the first process and the second process from the information obtained, it becomes possible to analyze the overall operation when the processes on each node advance the processing in a distributed system. Technology is disclosed.

  Further, Patent Document 3 provides each task with an execution order constraint between procedures obtained from the inter-procedure execution order constraint for each job in each task and the inter-procedure dependency, and this execution order for each task. A technique is disclosed that can efficiently and reliably obtain the same result as when a job is sequentially executed even when a plurality of jobs are simultaneously executed by performing execution control based on constraints.

  Furthermore, in Patent Document 4, in a real-time system using a parallel program, the execution time is calculated based on the time information and the determination condition obtained by the time measurement routine during execution of the parallel program, and the time constraint condition is verified. With this configuration, a technique is disclosed that can accurately and efficiently verify whether the time constraint set as the required specification is satisfied.

  However, none of the techniques disclosed in Patent Documents 2 to 4 relate to system monitoring.

JP 2005-327261 A JP 07-152704 A Japanese Patent Laid-Open No. 09-114693 Japanese Patent Laid-Open No. 11-242614

  As a method of monitoring whether the system is operating normally, collect trace data related to system operation during normal operation of the system, analyze the process execution status during normal operation based on the collected trace data, A method of monitoring the operation of the system during operation using the analyzed information is conceivable.

In such a method, when collecting trace data, an interface for collecting trace data is mounted on the board, and the trace data is extracted to an external device via this interface, and the extracted trace data is outside the system. Although an analysis method can be considered, it is not preferable to apply it to a system that is actually operated from the viewpoint of cost and security.
Further, as the number of CPUs increases due to the increase in the number of CPUs in the future, the amount of information to be collected increases, and a huge bandwidth is required for an interface for extracting the trace data to the outside. For this reason, it may be difficult to realize the above method.
On the other hand, no monitoring method has been proposed in which high-precision analysis is performed based on the process scheduling status in an OS (Operating System) during normal operation of the system, and all processes from analysis to monitoring are automated.

  As described above, the in-vehicle embedded system has a problem that it is necessary to monitor whether the system is operating normally.

  An object of the present invention is to solve the above-described problems, and to provide an information processing system, a monitoring method, and a monitoring program capable of monitoring whether the system is operating normally. is there.

  An information processing system according to the present invention is an information processing system including a first CPU that executes a plurality of processes and a second CPU that monitors a plurality of processes executed by the first CPU. An analysis unit that operates in the second CPU and extracts a dependency relationship among the plurality of processes based on an execution status of the plurality of processes; and an operation unit that operates in the second CPU and the analysis unit And a monitoring unit that detects abnormalities in the operation of the plurality of processes based on the dependency relationship extracted by the above.

  A monitoring method according to the present invention is a monitoring method for monitoring a plurality of processes executed by a first CPU from a second CPU, and based on the execution status of the plurality of processes, And a step of detecting an abnormal operation of the plurality of processes based on the dependency when the plurality of processes are monitored.

  A monitoring program according to the present invention is a monitoring method for monitoring a plurality of processes executed by a first CPU from a second CPU, and based on the execution status of the plurality of processes, For causing the second CPU to execute a step of extracting a dependency relationship between the plurality of processes and a step of detecting an abnormal operation of the plurality of processes based on the dependency relationship when monitoring the plurality of processes. It is.

  The present invention can provide an information processing system, a monitoring method, and a monitoring program that can monitor whether the system is operating normally.

It is a conceptual diagram which shows the outline | summary of the analysis monitoring system concerning embodiment of this invention. It is a block diagram of the analysis monitoring system concerning embodiment of this invention. It is a conceptual diagram which shows the element which operate | moves in the analysis phase of the analysis monitoring system concerning embodiment of this invention. It is a conceptual diagram which shows the element which operate | moves in the monitoring phase of the analysis monitoring system concerning embodiment of this invention. It is a figure which shows the process outline | summary in the analysis phase concerning embodiment of this invention. It is a flowchart of the production | generation process of the temporary dependence tree data concerning embodiment of this invention. It is a figure which shows an example of the dependency tree data concerning embodiment of this invention. It is a flowchart of the production | generation process of the dependency tree data concerning embodiment of this invention. It is a figure which shows the process outline | summary in the monitoring phase concerning embodiment of this invention. It is a figure which shows the process outline | summary in the monitoring phase concerning embodiment of this invention. It is a figure which shows the process outline | summary in the monitoring phase concerning embodiment of this invention. It is a figure which shows the process outline | summary in the monitoring phase concerning embodiment of this invention. It is a figure which shows the process outline | summary in the monitoring phase concerning embodiment of this invention. It is a flowchart of the monitoring process concerning embodiment of this invention.

Embodiments of the present invention will be described below with reference to the drawings.
First, with reference to FIG. 1, the outline | summary of the analysis monitoring system concerning this Embodiment is demonstrated. FIG. 1 is a conceptual diagram showing an outline of the analysis monitoring system according to the present embodiment.

The analysis monitoring system includes a CPU 80 that executes processes 70, 71, and 72, and a CPU 81 that operates the analysis unit 60 and the monitoring unit 61.
Based on the execution status of the processes 70, 71, and 72, the analysis unit 60 extracts a dependency relationship between the processes 70, 71, and 72.
When monitoring the processes 70, 71, 72, the monitoring unit 61 detects an abnormality in the operation of the processes 70, 71, 72 based on the dependency relationship extracted by the analysis unit 60.
Processes 70, 71, and 72 operate arbitrary programs, and execute predetermined processing. The number of processes is not limited to the number exemplified in this embodiment.
Each of the CPUs 80 and 81 may be a single core CPU or a CPU core included in a multi-core CPU. Note that the number of CPUs is not limited to the number exemplified in the present embodiment.

Subsequently, an outline of processing of the analysis monitoring system according to the present embodiment will be described.
First, the analysis unit 60 extracts the dependency between the processes 70, 71, 72 based on the execution status of the processes 70, 71, 72.
Next, when monitoring the processes 70, 71, 72, the monitoring unit 61 acquires the dependency relationship extracted by the analysis unit 60, and the abnormal operation of the processes 70, 71, 72 based on the acquired dependency relationship Is detected.

  Next, the analysis monitoring system according to the present embodiment will be described in detail with reference to FIGS. FIG. 2 is a configuration diagram of the analysis monitoring system according to the present embodiment, and FIGS. 3 and 4 are conceptual diagrams of the analysis monitoring system according to the present embodiment.

  As shown in FIG. 1, the analysis monitoring system includes a scheduling execution unit 1, an execution dependency detection unit 2, a dependency data shaping unit 3, a monitoring execution unit 4, temporary dependency tree data 5, and dependency tree data 6. 2 and 3, the analysis monitoring system includes CPUs 40, 41, and 42 that operate the scheduling execution unit 1, a CPU 43 that operates the execution dependency detection unit 2 and the dependency data shaping unit 3, and a monitoring execution unit 4. A CPU 44 to be operated and a DRAM (Dynamic Random Access Memory) 50 for storing the temporary dependency tree data 5 and the dependency tree data 6 are provided. The CPUs 40, 41, and 42 operate a plurality of arbitrary processes, but the illustration of these processes is omitted. These processes are included in the scheduling execution unit 1. The CPU 43 that operates the execution dependency detection unit 2 and the dependency data shaping unit 3 and the CPU 44 that operates the monitoring execution unit 4 may be the same CPU.

  The scheduling execution unit 1 schedules a plurality of processes and executes these processes. The scheduling execution unit 1 controls an execution queue (not shown) that is a list of processes that are in an executable state by scheduling. The scheduling execution unit 1 is configured by an OS scheduler, for example.

The execution dependency detection unit 2 extracts the dependency between the plurality of processes based on the execution status of the plurality of processes. Further, the execution dependency detecting unit 2 generates temporary dependency tree data 5 indicating the extracted dependency relationship, and stores the generated temporary dependency tree data 5 in the DRAM 50.
The dependency data shaping unit 3 acquires the temporary dependency tree data 5 stored in the DRAM 50 and shapes the acquired temporary dependency tree data 5 to generate dependency tree data 6. Then, the dependency data shaping unit 3 stores the generated dependency tree data 6 in the DRAM 50. The execution dependency detection unit 2 and the dependency data shaping unit 3 function as an analysis unit.

The monitoring execution unit 4 detects abnormal operation of a plurality of processes based on the dependency tree data 6 stored in the DRAM 50. The monitoring execution unit 4 functions as a monitoring unit.
The execution dependency detection unit 2, the dependency data shaping unit 3, and the monitoring execution unit 4 are implemented as part of functions in the OS, for example.

Each of the CPUs 40, 41, 42, 43, and 44 may be a single core CPU or a CPU core included in a multi-core CPU. Note that the number of CPUs is not limited to the number exemplified in the present embodiment.
The DRAM 50 is any storage device that can store or acquire the temporary dependency tree data 5 or the dependency tree data 6 at a speed that does not affect the processing performed in the execution dependency detection unit 2 or the execution dependency detection unit 2. There may be. The DRAM 50 corresponds to a storage unit.

  The analysis and monitoring system according to the present embodiment is divided into two phases. The two-stage phase is divided into an analysis phase in which normal operation of the system is analyzed to generate dependency tree data 6 and a monitoring phase in which the system is monitored based on the generated dependency tree data 6 and abnormal operation is detected. An overview of each phase according to the present embodiment will be described with reference to FIGS.

In the analysis phase, as shown in FIG. 3, the execution dependence detection unit 2 monitors the execution time of the interrupt generation, process generation, entry / exit of the execution queue, and processing executed by the process in the scheduling execution unit 1. Temporary dependency tree data 5 indicating the dependency between the two is generated.
The dependency data shaping unit 3 manages the dependency tree data 6, acquires the temporary dependency tree data 5 generated by the execution dependency detection unit 2, and generates dependency tree data 6 from the acquired temporary dependency tree data 5. The dependency tree data 6 is stored in the DRAM 50 so as not to overlap. In addition, it is preferable that the dependency data shaping unit 3 sets the execution priority lower than that of the execution dependency detection unit 2 so that the dependency data shaping unit 3 operates between the processes of the execution dependency detection unit 2 that needs to operate with high responsiveness. .

  As described above, in the analysis phase, the execution dependency detection unit 2 operated by the CPU 43 different from the CPUs 40, 41, and 42 that execute the process monitors the scheduling status of each process during normal operation, thereby Dependency tree data 6 is generated that indicates the dependency between processes.

  Further, the DRAM 50 is provided with a FIFO (First In First Out) buffer for transferring the dependency tree data 5 from the execution dependency detection unit 2 to the dependency data shaping unit 3, so that the dependency data is as long as the capacity of the DRAM 50 permits. The operation of the shaping unit 3 can be delayed. Furthermore, since the execution dependency detection unit 2 and the dependency data shaping unit 3 can be operated in parallel, a plurality of CPUs 43 are used, and the execution dependency detection unit 2 and the dependency data shaping unit 3 are operated in parallel by different CPUs. Speed up is possible.

In the monitoring phase, as shown in FIG. 4, the monitoring execution unit 4 monitors the generation of interrupts in the scheduling execution unit 1, the entry / exit of the execution queue, and the execution time of process processing, thereby generating dependency tree data generated in the analysis phase. It is determined whether or not the operation corresponding to 6 is performed. The monitoring execution unit 4 detects an abnormality by detecting a wake-up of a process or a delay in execution time at a timing not registered in the dependency tree data 6.
As described above, in the monitoring phase, in other words, during the operation operation, the monitoring execution unit 4 monitors the operation of the scheduling execution unit 1 based on the dependency tree data 6 during the normal operation, thereby detecting an abnormal operation. .

  Next, details of the analysis phase according to the present embodiment will be described with reference to FIGS. As shown in FIG. 3, the purpose of the analysis phase is to monitor and analyze the operation in the scheduling execution unit 1 and shape the temporary dependency tree data 5 generated by the analysis, thereby making the dependency tree data 6 available in the monitoring phase. Is to generate An outline of this series of processing is shown in FIG.

First, the process in the execution dependence detection part 2 is demonstrated.
FIG. 5A is a diagram illustrating an example of processing of a process executed in the scheduling execution unit 1. Here, it is assumed that the processes from the wake-up to the suspension of each process executed in the scheduling execution unit 1 are one section 90, 91, 92, 93, 94. Each process in FIG. 5A becomes executable at the timing of “wake-up” in each of the sections 90, 91, 92, 93, and 94, and is registered in the execution queue. Further, sections from “execution start” to “pause” in each of the sections 90, 91, 92, 93, and 94 indicate sections in which the execution right is given to the process and the process executes processing in each section.

First, the execution dependency detection unit 2 monitors the execution queue, and an interrupt is generated in the scheduling execution 1, and the registration of the wake-up process A in the execution queue is detected according to the generated interrupt. Thereby, the execution dependence detecting unit 2 detects the section 90.
Next, the execution dependence detecting unit 2 similarly monitors the execution queue to detect the sections 91 and 92 that have been started during the execution of the section 90. In this case, the execution dependence detector 2 determines that the detected sections 91 and 92 depend on the section 90 being executed, that is, the detected sections 91 and 92 are mutually dependent on the section 90 being executed. to decide. The reason for this is that, in such a case, the sections 91 and 92 are generally executed by triggering the processing in the section 90 of the process A, for example, processing such as unlocking or sending a message between processes. is there. Then, the execution dependency detecting unit 2 generates temporary dependency tree data 5 as shown in FIG. 5B by creating a branch indicating a dependency relationship between the section 90 and the sections 91 and 92. .

  Here, in this embodiment, since there are a plurality of CPUs 40, 41, and 42 that execute processes, there is a possibility that there are sections that are being executed in addition to the section 90. Since the OS recognizes the process that sent the message (process A in this case), it is possible to identify the dependent section by making a determination based on the OS.

Thereafter, similarly, when all the sections 90, 91, 92, 93, and 94 until the dependency relationship is not detected are executed, one temporary dependency tree data 5 as shown in FIG. 5B is generated. Is done.
Then, the execution dependency detection unit 2 stores the generated temporary dependency tree data 5 in the FIFO buffer of the DRAM 50 in order to pass it to the dependency data shaping unit 3.

  Next, with reference to FIG. 6, the details of the generation process of the temporary dependency tree data 5 described above will be described. FIG. 6 is a flowchart of the generation process of the temporary dependency tree data 5 according to the present embodiment. In FIG. 6, S <b> 10 to S <b> 14 indicate processing in the scheduling execution unit 1, and S <b> 100 to S <b> 106 indicate processing in the execution dependency detection unit 2.

First, the scheduling execution unit 1 registers a process in an executable state in the execution queue (S10).
The execution dependency detection unit 2 monitors the execution queue and detects registration of the process that has entered the executable state in the execution queue. As a result, the section where execution has been started is detected, and a currently-created dependency tree in which the detected section is registered as the head is created (S100). Here, the dependency tree being created is a tree in which the execution of each section registered in the dependency tree is not completed. Completion of execution of a section is determined to be complete when the process targeted by that section is removed from the execution queue.
Next, the scheduling execution unit 1 grants an execution right to the executable process registered in the execution queue, executes the process, and executes the process in the currently executing section (S11).

Next, when the scheduling execution unit 1 shifts the new process to an executable state and registers it in the execution queue during execution of the process (S12), the execution dependency detection unit 2 determines the registration factor (S101). .
If the new process registration factor is the activation factor, the execution dependency detection unit 2 creates a currently created dependency tree registered with the section detected in the registration of the execution queue as the head (S103). Here, the activation factor is a registration factor of a process that determines that the execution of a new series of sections that can be linked to one by dependency is started. In this embodiment, for example, an external interrupt, a process generation system call, or the like is used as an activation factor. That is, the registration of the execution queue in step S10 is also due to the activation factor.

On the other hand, when the registration factor of the new process is other than the activation factor, the execution dependency detection unit 2 determines that the process is awakened according to the processing in the currently executing section, and determines the dependency relationship in the currently executing section. The detected section is registered in the creating dependency tree (S102). That is, the execution dependency detection unit 2 determines that the detected section has a dependency relationship with the currently executing section.
Next, when the scheduling execution unit 1 finishes executing the process currently being executed (S13), the execution dependency detection unit 2 registers the execution time of the section that has been executed (S104). Here, the execution time is the execution time of the process in the section. The execution time is not limited to this. For example, the execution time of the section may be registered.

Next, the execution dependency detection unit 2 determines whether all sections registered in the currently created dependency tree have been executed at this stage, that is, a section that depends on the section that has been executed is included in the currently created dependency tree. It is determined whether it is registered and the corresponding process is registered in the execution queue (S105).
When execution of all sections is completed, the execution dependency detection unit 2 determines that the generation of the dependency dependency tree has been completed, and passes it to the dependency data shaping unit 3 as temporary dependency tree data 5 Then, it is stored in the FIFO buffer of the DRAM 50 (S106).
If the execution of all sections has not been completed, the scheduling execution unit 1 executes the next process registered in the execution queue (S14, S11).

Next, processing in the dependency data shaping unit 3 will be described.
The dependency data shaping unit 3 generates dependency tree data 6 used in the monitoring phase as illustrated in FIG. 5C based on the temporary dependency tree data 5 as illustrated in FIG. Here, the processing performed by the dependency data shaping unit 3 is integration of divided sections by the time division scheduling method, detection of loop processing, and avoiding duplicate registration with existing dependency tree data 6.

  As illustrated in FIG. 7, the dependency tree data 6 generated by the dependency data shaping unit 3 includes information indicating an activation factor and a series of sections that are executed while having a dependency. In other words, a series of sections is a collection of sections that can be linked together by dependency. One item indicating one section includes a process ID uniquely determined for each process and a section ID including a section number uniquely determined for each section, a dependent section ID, an execution start program counter value, and an execution time. They are arranged in the order registered in the execution queue. That is, the execution dependence detection unit 2 generates temporary dependence tree data 5 including these pieces of information. Here, the execution start program counter value is a program counter value at which the process starts processing in the section, and is used for detection of loop processing and duplication detection of the dependency tree data 6.

  Next, details of the generation process of the dependency tree data 6 described above will be described with reference to FIG. FIG. 8 is a flowchart of the generation process of the dependency tree data 6 according to the present embodiment.

First, the dependency data shaping unit 3 acquires one temporary dependency tree data 5 from the FIFO buffer in which the temporary dependency tree data 5 is recorded (S200).
Next, when there are two or more sections in the temporary dependency tree data 5 that have the same section ID, the dependent data shaping unit 3 sets them as one section (S201, S202). When the right to execute a process is switched at a predetermined time as in the time-sharing scheduling method, what should essentially become one section may be divided in the temporary dependency tree data 5. In such a case, since two or more sections having the same dependency destination section ID are created in the temporary dependency tree data 5, these sections are combined into one. In other words, when the execution dependency detection unit 2 detects registration of the execution queue due to switching of the time-sharing scheduling, if the same process as the process being executed is registered in the execution queue, it is newly added to the created dependency tree. As the dependency destination section of the section, the same section as the dependency destination section of the section being executed is stored. Note that the value of the first section of the temporary dependency tree data 5 among the two or more sections to be combined is used as the execution start program counter value of the section generated by the combination, and the combination of 2 is used for the execution time. Use the value obtained by adding the execution times of two or more sections.

  Next, the dependent data shaping unit 3 detects a loop process (S203). Here, the dependency data shaping unit 3 collects the portions in which the same number sequence is repeated into one loop process when the execution start program counter values of the sections are arranged in the order registered in the execution queue. An example is shown in the dependency tree data 6 starting from the section “process B-1” with “external interrupt A” in FIG. 7 as the activation factor. Here, it is shown that the processing in the sections “process B-2” and “process A-1” loops. Instead of the section, the information indicating the loop may store information about the loop such as the number of loops.

Next, the dependency data shaping unit 3 searches whether there is similar dependency tree data 6 in the dependency tree data 6 stored in the DRAM 50 (S204, S205).
When there is similar dependency tree data 6, specifically, the execution start program counter value of each section is the same, that is, has the same dependency tree shape, the execution time of each section is close, and the difference is a predetermined threshold value. If there is dependency tree data 6 that falls within the dependency tree, the dependency data shaping unit 3 integrates these two dependency tree data 6 and stores them as one dependency tree data 6 in the DRAM 50 (S206). At this time, the execution time of each section is calculated as an average value, and the calculated average value is stored as the execution time.
When the same dependency tree data 6 does not exist, the dependency data shaping unit 3 stores the generated dependency tree data 6 in the DRAM 50.

In this way, the dependency data shaping unit 3 can reduce the amount of data stored in the DRAM 50 by integrating sections, detecting loop processing, and avoiding duplicate registration with existing dependency tree data 6. .
The dependency tree data 6 is not limited to the dynamic analysis as described above, but may be created by the user and directly registered in the DRAM 50.

  Next, details of the monitoring phase according to the present embodiment will be described with reference to FIGS. As shown in FIG. 4, the purpose of the monitoring phase is to detect an abnormality by comparing the operation in the scheduling execution unit 1 with the normal operation shown in the dependency tree data 6. In the monitoring phase, the dependency tree data 6 suitable for the operation in the scheduling execution unit 1 is selected from the dependency tree data 6 generated in the analysis phase according to the activation factor and the section to be executed. When it is detected that there is no corresponding dependency tree data 6, it is determined that there is an abnormality. An outline of a series of monitoring processes is shown in FIGS. Here, a case where the dependency tree data 6 illustrated in FIG. 7 is stored in the DRAM 50 is illustrated.

First, FIG. 9 shows a situation where the external interrupt A has occurred as an activation factor. The dependency tree data 6 is collectively stored in the DRAM 50 for each activation factor, and the monitoring execution unit 4 indicates the external interrupt A as the activation factor as a monitoring target dependency tree list in which sections that may be executed are stored. Dependent dependency tree data 6 is selected. Here, the monitoring target dependency tree list is the dependency tree data 6 in the dependency tree data 6 to which a series of sections to be executed from now on may match.
Next, FIG. 10 shows a situation in which the process A is registered in the execution queue due to the activation factor, and the execution of the section “process A-1” is started. Therefore, the monitoring execution unit 4 leaves only the monitoring target dependency tree that indicates the section that is first executed by the process A in the monitoring target dependency tree list as the monitoring target dependency tree list.

  Next, in FIG. 11, the process C is woken up and registered in the execution queue by the execution of the process A, and the external interrupt B is generated as an activation factor, and the process D is registered in the execution queue accordingly. Is shown. The monitoring execution unit 4 selects a new monitoring target dependency tree list corresponding to the external interrupt B and the process D. As described above, when a new monitoring target is generated by the external interrupt B that is the activation factor, the monitoring execution unit 4 generates a new monitoring target existing tree list according to the new monitoring target. Further, the monitoring execution unit 4 associates a process registered in the execution queue with any monitoring target dependency tree list in order to clarify which monitoring target dependency tree list the process relates to.

Here, the monitoring target dependency tree list that is associated with the process registered in the execution queue for reasons other than the activation factor is the monitoring target dependency that is associated with the process that is currently executing when the process is registered in the execution queue. It is a tree list. This is because the process registered in the execution queue is registered in the execution queue with the start of the execution of the section depending on the currently executing section. If multiple processes are executed in parallel, as already mentioned, the OS recognizes the process that wakes up the process registered in the execution queue, so the process can be identified based on it. It is.
On the other hand, when the execution of the section at the head of the dependency tree data 6 is started by the activation factor and the process is registered in the execution queue, the monitoring generated by the activation factor is independent of the currently executing process. Corresponds to the target dependency tree list.

  Next, when the normal operation is performed, the process C that executes the process in the section “process C-1” depending on the section “process A-1” currently being executed is registered in the execution queue. It becomes a process. In FIG. 11, since the process C is registered in the execution queue, the monitoring execution unit 4 determines that the operation is normal.

Next, FIG. 12 shows a state where the execution of the process C is started in the scheduling execution unit 1. The monitoring execution unit 4 leaves only the monitoring target dependency tree in which the section next to the previously executed section “process A-1” indicates the process C as a monitoring target dependency tree list. Also, sections that have been executed are excluded from monitoring targets. In normal operation, the process proceeds until all sections are excluded.
On the other hand, as shown in FIG. 13, the monitoring execution unit 4 determines whether the process not registered in the monitoring target dependency tree list is registered in the execution queue or compared with the execution time registered in the monitoring target dependency tree list. If the processing execution time in a section is long, it is detected as abnormal. When the monitoring execution unit 4 detects an abnormality, the monitoring execution unit 4 stores information on an abnormality such as a process not included in the monitoring target dependency tree list or the monitoring target dependency tree list registered in the execution queue, for example, in a memory or HDD (Hard Disc The output may be output to an arbitrary storage device (not shown) such as a drive) or an arbitrary display device (not shown) such as a CRT (Cathode Ray Tube), a plasma display, or a liquid crystal display.

  Next, details of the monitoring process described above will be described with reference to FIG. FIG. 14 is a flowchart of the monitoring process according to the present embodiment. In FIG. 14, S30 to S34 indicate processing in the scheduling execution unit 1, and S300 to S306 indicate processing in the execution dependency detection unit 2.

First, when an activation factor occurs, the scheduling execution unit 1 registers a process that has become executable according to the activation factor in the execution queue (S30).
When the process is registered in the execution queue according to the activation factor, the monitoring execution unit 4 generates a monitoring target dependency tree list (S300). At this time, the monitoring execution unit 4 generates the monitoring target dependency tree list by selecting the dependency tree data 6 that matches the activation factor and the process registered in the execution queue according to the activation factor. To do.
Next, the scheduling execution unit 1 grants an execution right to the executable process registered in the execution queue, executes the process, and executes the process in the section (S31).

Next, when the scheduling execution unit 1 shifts the new process to an executable state and registers it in the execution queue during execution of the process (S32), the monitoring execution unit 4 determines the registration factor (S301).
When the registration factor of the new process is the activation factor, the monitoring execution unit 4 selects the new process and the dependency tree data 6 that matches the activation factor, and creates a monitoring target dependency tree list (S302).
On the other hand, when the registration factor of the new process is other than the activation factor, the monitoring execution unit 4 searches the monitoring target dependency tree list associated with the process being executed, It is determined whether there is a monitoring target dependency tree in which a process indicated by a section having a dependency relationship with the section being executed matches a new process (S303).

If there is a match, or if the process currently being executed and the process registered in the execution queue are the same, that is, if switching by time-sharing scheduling is performed, the monitoring execution unit 4 determines that it is normal .
If there is no match, the monitoring execution unit 4 determines that an abnormality has occurred.
Here, it is possible to detect an abnormality with high accuracy by further determining whether or not the execution start program counter value of the monitoring target dependency tree matches the execution start program counter value of the processing in the section executed by the process execution. It may be.

  The monitoring execution unit 4 also exceeds the execution time of the currently executing process exceeding the longest execution time corresponding to the currently executing section of the monitoring target dependency tree remaining in the monitoring target dependency tree list. It is determined that an abnormality has occurred (S304). Here, the reason why it is the longest is that when a plurality of monitoring target dependency trees remain, each monitoring target dependency tree has an execution time corresponding to the section currently being executed. The execution time may be determined that an abnormality has occurred when the difference between the execution time of the monitoring target dependency tree and the execution time of the currently executing process exceeds a predetermined threshold.

Next, when the scheduling execution unit 1 finishes executing the currently executing process (S33), the monitoring execution unit 4 determines that the monitoring target dependency tree list associated with the process has only one dependency tree. It is determined whether all the sections are executed (S305). Note that the end of process execution can be determined by detecting the state in which the process is removed from the execution queue.
When the monitoring target dependency tree list is only one dependency tree and all the sections are executed, the monitoring execution unit 4 determines that the execution of the series of sections has been normally completed. .
In other cases, when the scheduling execution unit 1 executes the next process registered in the execution queue (S34), the monitoring execution unit 4 monitors the monitoring target that does not conform to the process from the monitoring target dependency tree list. The dependency tree is deleted (S306), and the process is continued.

As described above, according to the present embodiment, based on the execution status of a plurality of processes, a dependency relationship between a plurality of processes is extracted, and a plurality of processes are extracted based on the extracted dependency relationship. By detecting an abnormal operation, it is possible to monitor whether the system is operating normally.
Furthermore, according to the present embodiment, it is possible to monitor the system in an operating environment without requiring an external device or an external interface. This also eliminates an increase in cost and security problems caused by mounting external devices and external interfaces.
Furthermore, according to this embodiment, the system is analyzed to automatically generate monitoring data, and the system is monitored using the automatically generated monitoring data, so that the system analysis to the monitoring can be performed. All can be done automatically.
Further, according to the present embodiment, when an abnormality occurs, it is possible to grasp a series of system operations up to immediately before by referring to information output by the monitoring execution unit.
Furthermore, according to the present embodiment, analysis and monitoring are performed based on the scheduling status of the process, so that highly accurate monitoring can be performed.

Note that the present invention is not limited to the above-described embodiment, and can be modified as appropriate without departing from the spirit of the present invention.
For example, in the present embodiment, the mutual dependency between sections, which is the process from the wake-up of the process to the suspension, is extracted. However, the present invention is not limited to this as long as the dependency is related to the process. For example, mutual dependency relations may be managed for processes executed by processes in sections.

The analysis and monitoring system according to the present invention described above supplies a storage medium storing a program for realizing the functions of the above-described embodiments to the system or apparatus, and the computer or CPU, MPU (Micro Processing) included in the system or apparatus. Unit) can be configured by executing this program.
In addition, this program can be stored in various types of storage media and can be transmitted via a communication medium. Here, examples of the storage medium include a flexible disk, a hard disk, a magnetic disk, a magneto-optical disk, a CD-ROM (Compact Disc Read Only Memory), a DVD (Digital Versatile Disc), a BD (Blu-ray Disc), and a ROM ( A read only memory (RAM) cartridge, a battery-backed RAM (Random Access Memory) memory cartridge, a flash memory cartridge, and a nonvolatile RAM cartridge are included. The communication medium includes a telephone line wired communication medium and a microwave line wireless communication medium, and includes the Internet.

Further, when the computer executes the program that realizes the functions of the above-described embodiment, not only the functions of the above-described embodiment are realized, but also the computer is operating on the basis of the instructions of this program. The case where the functions of the above-described embodiment are realized in cooperation with the OS or application software is also included in the embodiment of the invention.
Further, when the functions of the above-described embodiment are realized by performing all or part of the processing of the program by a function expansion board inserted into the computer or a function expansion unit connected to the computer, the present invention may be implemented. It is included in the form.

DESCRIPTION OF SYMBOLS 1 Scheduling execution part 2 Execution dependence detection part 3 Dependent data shaping part 4 Monitoring execution part 5 Temporary dependence tree data 6 Dependency tree data 40, 41, 42, 43, 44, 80, 81 CPU
50 DRAM
60 Analysis Unit 61 Monitoring Unit 70, 71, 72 Process

Claims (9)

An information processing system comprising: a first CPU that executes a plurality of processes; and a second CPU that monitors a plurality of processes executed by the first CPU,
Based on the scheduling status of the plurality of processes based on the scheduling status of the plurality of processes , depending on the execution of an arbitrary section as a dependency relationship between the sections, which are processes from wakeup to suspension of the plurality of processes An analysis unit that extracts sections to be executed ;
A section that operates in the second CPU and that is executed according to execution of an arbitrary section extracted by the analysis unit based on the dependency relationship extracted by the analysis unit, and according to execution of the arbitrary section An information processing system comprising: a monitoring unit that detects an abnormality in operation of the plurality of processes by comparing whether or not the sections to be executed match . The analysis unit further extracts an execution time of a process executed by the process in the section,
The information processing system according to claim 1 , wherein the monitoring unit detects the abnormality by comparing whether or not the execution times coincide with each other. The information processing system further includes a storage unit that stores the content extracted by the analysis unit,
The analysis unit stores the extracted content as dependency relationship information in the storage unit in association with each of a series of sections that can be connected to one by the dependency relationship,
The monitoring unit information processing system according to claim 1 or 2, based on the dependency information stored in the storage unit, performs the comparison. The information processing system according to claim 3 , wherein the analysis unit stores the same dependency relationship information in the storage unit so as not to include two pieces of dependency information. The analysis unit, an information processing system according to claim 3 or 4 for storing said series of sections, time division dependency information the divided sections are summarized in one section by the scheduling method. The analysis unit further extracts an execution start program counter value of the process executed by the process in the section, and the dependency in which the change of the same execution start program counter value is repeated is combined into one in the series of sections. the information processing system according to any one of claims 3 to 5 for storing the relationship information. The information processing system according to claim 6 , wherein the monitoring unit further detects the abnormality by comparing whether or not the execution start program counter value also matches. A monitoring method for monitoring a plurality of processes executed by a first CPU from a second CPU,
Extracting a section to be executed in response to execution of an arbitrary section as a dependency relationship between sections, which are processes from wakeup to suspension of the plurality of processes, based on scheduling states of the plurality of processes; ,
When monitoring the plurality of processes, the section executed in response to the execution of the extracted arbitrary section matches the section executed in response to the execution of the arbitrary section based on the dependency relationship. And a step of detecting an abnormal operation of the plurality of processes by comparing whether or not to perform the monitoring. A monitoring program for monitoring a plurality of processes executed by a first CPU from a second CPU,
Extracting a section to be executed in response to execution of an arbitrary section as a dependency relationship between sections, which are processes from wakeup to suspension of the plurality of processes, based on scheduling states of the plurality of processes; ,
When monitoring the plurality of processes, the section executed in response to the execution of the extracted arbitrary section matches the section executed in response to the execution of the arbitrary section based on the dependency relationship. A monitoring program that causes the second CPU to execute a step of detecting an abnormality in the operation of the plurality of processes by comparing whether or not to do so .

Images