JP5974856B2 - Sampling program, sampling method and information processing apparatus - Google Patents

Description

  The present invention relates to a sampling program, a sampling method, and an information processing apparatus that collect computer data.

  In order to optimize a program, a sampling method is known in which a program mounted on a computer is executed, operation information such as an operation function and an execution time called at the time of program execution is periodically sampled, and data is collected. . The sampling of operation information is suitable for detecting an operation function or the like having a relatively large number of calls at the time of program execution because the overhead (processing load) at the time of information acquisition is small. The operation information collected by sampling includes, for example, identification information (PID) of a process executed at the time of sampling, an instruction address, and the like.

  In the sampling, the collected operation information is accumulated in the sampling order, for example, in a memory of a computer or a storage device connected to the computer.

  In addition, the following patent documents exist as prior art documents in which technologies related to the technologies described in this specification are described.

Japanese Patent Laid-Open No. 2000-10828 JP 2006-202134 A

For example, it has been known that the processing power of a computer is improved by technologies such as a multiprocessor having a plurality of processors (CPU: Central Processing Unit) and a multicore processor in which a plurality of cores (Core) are incorporated in a single CPU package. It has been.

  In the optimization of a program executed on such a computer, for example, operation information is sampled for each processor and each core, and the collected data is accumulated. For this reason, in the conventional sampling, the amount of data collected in proportion to the number of cores and the number of processors increases, which may exceed the capacity of a memory, a storage device, or the like that accumulates data.

  The same situation can occur even when the execution time of the application program executed by the computer is long.

  In one aspect, an object of the present invention is to provide a technique for reducing the amount of accumulated data in sampling of operation information.

The above technique is exemplified by the configuration of the following sampling program. In other words, the sampling program compares the sampling information related to the operation of the target program collected at a predetermined sampling period during the execution period of the target program to be sampled with a computer, and compares the sampling information adjacent in time series. A detection step for detecting that sampling information having common elements is continuous in time series, and a counting step for counting the number of consecutive samplings of the sampling information detected in the detection step to be continuous in time series; And a recording step of recording the sampling information detected to be continuous in time series in the detection step and the continuous sampling count counted in the counting step in association with each other.

  According to the above sampling program, it is detected that sampling information having a common element is continuous in time series, and the continuous sampling information and the continuous sampling count are recorded in association with each other. Can be reduced.

  According to the above sampling program, it is possible to provide a technique for reducing the amount of accumulated data in sampling operation information.

It is a schematic block diagram of the sampling device of this embodiment. It is a figure which illustrates the sampling data by the conventional sampling method. It is a schematic block diagram of the hardware of a sampling device. It is a figure for demonstrating the function structure of a sampling apparatus. It is a figure which illustrates the record stored in sampling DB. It is a figure which illustrates the function information table stored in function information DB. It is a flowchart which illustrates the creation process of a function information table. It is a flowchart which illustrates the recording process of sampling data.

  Hereinafter, a sampling apparatus according to an embodiment will be described with reference to the drawings. The configuration of the following embodiment is an exemplification, and the sampling apparatus is not limited to the configuration of the embodiment.

  Hereinafter, the sampling apparatus will be described with reference to the drawings of FIGS.

〔Device configuration〕
FIG. 1 is a schematic configuration diagram of a sampling apparatus according to the present embodiment. The sampling apparatus 100 in FIG. 1 is an information processing apparatus that functions as a multiprocessor computer including a plurality of processors (CPUs: Central Processing Units), for example. The sampling apparatus 100 of FIG. 1 implement | achieves the multiprocessing system (multi processing system) which performs a some processing process in parallel by a some processor.

  The sampling apparatus 100 in FIG. 1 samples operation information together with execution of a target program. In sampling of operation information, the sampling apparatus 100 records a function being operated (hereinafter referred to as an operation function) at each sampling interval. In the sampling of operation information, the sampling device 100 performs data compression when data indicating sampling data in the same operation function is continuously sampled.

  In the following description, an example of processing executed by the sampling apparatus 100 having a plurality of processors will be described as an example of processing by a computer in which the sampling program of the present embodiment is installed. However, the computer on which the sampling program of the present embodiment is installed is not limited to the sampling apparatus 100.

As a computer on which the sampling program of this embodiment is mounted, for example, a computer using a multi-core processor in which a plurality of cores (Core) are incorporated in a single CPU package, or a computer having a plurality of multi-core processors. Good. Further, the computer on which the sampling program of the present embodiment is installed may be a supercomputer used in a technical field such as HPC (High-performance computing). A computer applicable to the present embodiment is a computer that samples operation information such as an operation function at the time of program execution and accumulates the collected sampling data in a predetermined storage device or the like in order to optimize the program. That's fine.

  The sampling device 100 in FIG. 1 is connected to the storage device 200. Note that the sampling device 100 may include the storage device 200. The sampling device 100 and the storage device 200 may be part of a cloud that is a group of computers on a network, for example. The network includes a public network such as the Internet, a local area network (LAN), or a wide area network (WAN).

  The storage device 200 is a storage device that includes a storage medium that stores various programs and various data. The storage device 200 is also called an external storage device. Examples of the storage device 200 include a solid state drive device and a hard disk drive device. The storage device 200 can include a portable recording medium such as a CD (Compact Disc) drive device, a DVD (Digital Versatile Disc) drive device, and a BD (Blu-ray Disc) drive device.

  The storage device 200 includes a sampling DB 210, an operation function DB 220, and a function information DB 230. The sampling DB 210, the operation function DB 220, and the function information DB 230 may each be included in separate storage devices.

  The sampling apparatus 100 in FIG. 1 collects information related to the operation of the target program at a predetermined sampling period during the execution period of the program to be sampled (hereinafter referred to as the target program). The sampling device 100 collects the collected sampling data and accumulates the collected sampling data in the storage device 200.

  The sampling period of sampling executed by the sampling apparatus 100 is arbitrary as long as it does not exceed the execution period of the target program. For example, time unit intervals such as 1 second, 10 milliseconds, and 100 microseconds can be exemplified. An appropriate period may be set according to the performance of the computer, the execution period of the target program, the number of samples, and the like.

  FIG. 2 illustrates sampling data collected by a conventional sampling method. The sampling data in FIG. 2 has a record for each sampling. In the sampling data of the example, “CPU number”, “PID”, and “instruction address” are collected and collected as information related to the operation of the target program. “CPU number”, “PID”, “instruction address”, and the like are examples of elements included in the sampling information. Here, the “CPU number” is the number of the CPU related to the execution of the program, the “PID” is the identification information of the process that is executing the program, and the “instruction address” is the information of the address position that is executing the instruction.

  Thus, in the conventional sampling method, sampling data including “CPU number”, “PID”, and “instruction address” is accumulated as a record for each sample until the execution period of the target program ends.

  For example, the amount of data collected by one sampling can be exemplified by the following equation (1).

Sampling data amount = 1 record data size x number of samples (execution period / sampling rate)
X number of CPUs x number of cores (1)
Therefore, it can be seen that the amount of data collected by one sampling increases in proportion to the record data size, the number of samples, the number of CPUs, and the number of cores according to the equation (1).

  The sampling apparatus 100 according to the present embodiment detects, for example, that sampling data indicating data in the same operation function continues from sampling data collected in time series. Then, the sampling apparatus 100 according to the present embodiment counts the number of consecutive samplings of sampling data indicating that the data is in the same operation function. The sampling apparatus 100 of this embodiment replaces recording individual sampling data one by one, sampling data indicating data in the same operation function, and sampling data indicating data in the same operation function Are recorded in association with the number of consecutive samplings. As a result, the accumulated data amount of data accumulated in the sampling apparatus 100 of the present embodiment is compressed, so that the collected sampling data amount is reduced.

  In addition, the sampling apparatus 100 according to the present embodiment specifies an operation function having a relatively high sampling frequency of the target program from the sampling data collected in time series. The identified operation function having a relatively high sampling frequency is held in a table format, for example. The sampling apparatus 100 according to the present embodiment reduces the processing time for detecting an operation function from the collected sampling data by holding an operation function having a relatively high sampling frequency in a table format.

  FIG. 3 illustrates a schematic configuration diagram of hardware of the sampling device 100. The sampling device 100 includes a plurality of CPUs 101 # 1-CPU101 # n, a main storage unit 102, an auxiliary storage unit 103, and a communication unit 104 connected to each other via a connection bus B1. In the sampling device 100, the CPU 101 # 1-CPU101 # n develops a program stored in the auxiliary storage unit 103 so as to be executable in the work area of the main storage unit 102, and controls peripheral devices through execution of the program. Thereby, the sampling apparatus 100 can implement | achieve the function corresponding to the predetermined objective. The main storage unit 102 and the auxiliary storage unit 103 are recording media that can be read by the sampling apparatus 100 that is a computer.

  CPU101 # 1-CPU101 # n are central processing units of the sampling device 100. CPU 101 # 1 -CPU 101 #n perform processing according to a program stored in auxiliary storage unit 103. The main storage unit 102 is a storage medium in which the CPU 101 # 1-CPU101 # n caches programs and data and develops a work area. The main storage unit 102 includes, for example, a RAM (Random Access Memory) and a ROM (Read Only Memory).

  The auxiliary storage unit 103 stores various programs and various data in a recording medium in a readable and writable manner. The auxiliary storage unit 103 stores an operating system (OS), various programs, various tables, and the like. The OS includes a communication interface program that exchanges data with an external device or the like connected via the communication unit 104. The external device or the like includes, for example, other information processing devices and storage devices connected via a network. The communication unit 104 is an interface with a network or the like, for example.

The auxiliary storage unit 103 is, for example, an EPROM (Erasable Programmable ROM), a solid state drive device, a hard disk drive (HDD, Hard Disc Drive) device, or the like. As the auxiliary storage unit 103, for example, a CD drive device, a DVD drive device, a BD drive device, or the like can be presented. Examples of the recording medium include a silicon disk including a nonvolatile semiconductor memory (flash memory), a hard disk, a CD, a DVD, a BD, and a USB (Universal Serial Bus) memory.

The sampling apparatus 100 illustrated in FIG. 3 may include an input unit that receives an operation instruction or the like from a user or the like, for example. Examples of such an input unit include an input device such as a keyboard and a pointing device. The sampling apparatus 100 may include an output unit that outputs data processed by the CPUs 101 # 1-CPU101 # n and data stored in the main storage unit 102. As such an input unit, a CRT (Cathode Ray Tube) display, an LCD (Liquid Crystal Display), a PDP (Plasma)
Examples include output devices such as a display panel, an EL (electroluminescence) panel, an organic EL panel, and a printer.

  The sampling apparatus 100 is illustrated in FIG. 1 together with the execution of the target program by the CPU 101 # 1-CPU101 # n reading various programs and various data stored in the auxiliary storage unit 103 into the main storage unit 102 and executing them. Realize each functional unit. Along with the execution of the target program, the sampling apparatus 100 includes a first sampling processing unit 110, an operation function specifying unit 120, a frequency calculation unit 130, a function information table creation unit 140, a second sampling processing unit 150, a continuous detection example illustrated in FIG. Unit 160 and collected data storage unit 170 are realized.

  Note that any one of the functional units may be included in another information processing apparatus. For example, the sampling device 100 includes a first sampling processing unit 110, a second sampling processing unit 150, a continuous detection unit 160, and a collected data storage unit 170, and is connected to a network. Then, the information processing apparatus including the operation function identification unit 120, the information processing apparatus including the frequency calculation unit 130, and the information processing apparatus including the function information table creation unit 140 are connected to the network. As described above, the sampling device 100 may be realized by distributing functional units to a plurality of information processing devices and executing the functional units. Since the sampling device 100 can be realized in the cloud that is a group of computers on the network, the processing load of each functional unit can be reduced.

[Function block configuration]
FIG. 4 illustrates a diagram for explaining the functional configuration of the sampling apparatus 100. The sampling apparatus 100 illustrated in FIG. 4 includes a first sampling processing unit 110, an operation function specifying unit 120, a frequency calculation unit 130, a function information table creation unit 140, a second sampling processing unit 150, a continuous detection unit 160, and collected data storage. Each functional unit of the unit 170 is included.

  The sampling device 100 illustrated in FIG. 4 includes a sampling DB 210, an operation function DB 220, and a function information DB 230 constructed in the storage device 200 as a storage destination of data to be referred to or managed by the above functional units. .

  The first sampling processing unit 110 executes the target program in a specific period, and collects information related to the operation of the target program collected at a predetermined sampling period during the specific period. The information collected by the first sampling processing unit 110 is stored in a predetermined area of the main storage unit 102, for example. The collected information is delivered to the action function identification unit 120 and the frequency calculation unit 130.

  The sampling data collected by the first sampling processing unit 110 is collected, for example, in the sampling data format illustrated in FIG. The sampling data collected by the first sampling processing unit 110 has a record for each sampling. The first sampling processing unit 110 collects and collects information such as “CPU number”, “PID”, and “instruction address” as information related to the operation of the target program.

  Note that the sampling period of sampling data executed by the first sampling processing unit 110 is set to a specific period shorter than the execution period of the target program, for example. The set specific period is stored in the auxiliary storage unit 103, for example. Further, the specific period may be set by an administrator of the sampling apparatus 100 or the like, for example, through an operation input such as a keyboard, or may be set via a network connected to the sampling apparatus 100. The first sampling processor 110 only needs to be able to collect sampling data in a specific period shorter than the execution period of the target program.

  The operation function specifying unit 120 specifies the operation function information associated with the PID and the instruction address with reference to the operation function DB 220 from the sampling data delivered from the first sampling processing unit 110. The action function information includes a program name corresponding to the PID, a function name of the action function, a start address and an end address of the action function. The start address is, for example, the instruction address of the first instruction of the function in the program. The end address is, for example, the instruction address of the instruction at the end of the function in the program. The action function specifying unit 120 stores the specified action function information in a predetermined area of the main storage unit 102, for example. The identified operation function information is delivered to the frequency calculation unit 130.

  Based on the sampling data delivered from the first sampling processing unit 110 and the action function information delivered from the action function specifying part 120, the frequency calculating part 130 totals the number of samples for each specified action function. The frequency calculation unit 130 arranges the identified operation functions in the order of the relatively large number of samples, and stores them in a predetermined area of the main storage unit 102. The operation functions arranged in the order of the relatively large number of samples are delivered to the function information table creation unit 140.

  The function information table creation unit 140 assigns function identification information that uniquely identifies the motion function to the motion function delivered from the frequency calculation unit 130. Then, the function information table creation unit 140 creates a function information table by associating the assigned function identification information with the action function name, the program name corresponding to the PID, the start address and the end address position of the action function. The created function information table is stored in a predetermined area of the main storage unit 102. Further, the function information table creation unit 140 stores the created function information table in the function information DB 230.

  The function information table creation unit 140 may generate a function information table for an operation function having a predetermined threshold value and having a sample number exceeding the predetermined threshold value, for example. Further, for example, the function information table creation unit 140 may generate a function information table for a predetermined number (for example, the top 20) operation functions. In addition, for example, the function information table creation unit 140 occupies a certain ratio (for example, 50% of the whole) among all the motion functions aggregated by the frequency calculation unit 130, and the function information table creation unit 140 An information table may be created.

  In creating the function information table, an appropriate function information table must be created according to the operating environment in which the target program is executed, the execution time of the target program, the sampling period, the storage capacity of the memory that stores the sampling data, etc. Criteria can be selected.

  The second sampling processing unit 150 collects information related to the operation of the target program at a predetermined sampling period during the execution period of the target program, and collects sampling data. The collected sampling data is collected, for example, in the sampling data format illustrated in FIG. The second sampling processing unit 150 temporarily stores the collected sampling data in a predetermined area of the main storage unit 102.

Further, the second sampling processing unit 150 identifies an operation function from the address in the program included in the sampling data, and the sampling data is one of the functions defined in the function information table generated by the function information table creation unit 140. Determine whether it is in the address range.

  When the sampling data is in the instruction address range of any function defined in the function information table, the second sampling processing unit 150 specifies function identification information corresponding to the sampling data from the function information table. The identified function identification information is selected as recording information to be accumulated as sampling data. The second sampling processing unit 150 stores the specified function identification information in a predetermined area of the main storage unit 102 and delivers it to the continuous detection unit 160 and the collected data storage unit 170.

  On the other hand, when the operation function identified from the sampling data is not included in the function information table, the second sampling processing unit 150 uses the address position information during the execution of the instruction included in the sampling data as the sampling data as in the conventional case. Select recording information to be stored. The address position information during instruction execution included in the sampling data is, for example, an “instruction address” illustrated in FIG.

  The second sampling processing unit 150 stores the address position information during the execution of the instruction included in the sampling data in a predetermined area of the main storage unit 102 and passes it to the continuous detection unit 160 and the collected data storage unit 170.

  The continuous detection unit 160 detects that the same operation function information or the same PID and instruction address are consecutive in the sampling time-series order for the sampling data collected by the second sampling processing unit 150. The continuous detection unit 160 compares the collected sampling data with the sampling data collected in the immediately preceding sampling period, and detects the sampling data in which the same operation function information or the same PID and instruction address are consecutive in time series order. To do.

  For example, the continuous detection unit 160 compares the “CPU number” and “PID” of the nth sampling data with the “CPU number” and “PID” of the (n−1) th sampling data. I do. As a result of the comparison, the continuous detection unit 160 can detect that the sampled data is continuous in time series when both of the “CPU number” and “PID” are the same.

  In addition, when the instruction address in the program included in the sampling data is within the range of any function in the function information table, the continuity detection unit 160 performs the same processing on the function identification information specified from the function information table. Just do it. For example, the continuous detection unit 160 may compare the function identification information of the nth sampling data with the function identification information of the (n−1) th sampling data. As a result of the comparison, the continuous detection unit 160 can detect that the sampled data is continuous in time series when the function identification information is the same.

  Further, when the continuous detection unit 160 detects continuous sampling of the same function identification information (or the same combination of PID and instruction address), for example, a continuous detection counter set in a predetermined area of the main storage unit 102 The count value is incremented by “1”. The continuous detection unit 160 keeps incrementing the count value of the continuous detection counter every time the same operation function information (or the combination of the same PID and instruction address) detects continuous sampling, thereby reducing the continuous number of sampling data. Can count.

  The continuous detection unit 160 stores, in a predetermined area of the main storage unit 102, the number of consecutive sampling data in which the same operation function information (or combination of the same PID and instruction address) counted is stored, and collects data storage Delivered to the unit 170.

  The collected data storage unit 170 stores the delivered sampling data in the sampling DB 210, for example, in the sampling data format illustrated in FIG. 5A.

[Database configuration]
FIG. 5A illustrates the sampling DB 210. In the sampling DB 210 of FIG. 5A, sampling data records created by the collected data storage unit 170 are stored. The sampling DB 210 has a record for each action function detected continuously. Each record of the sampling DB 210 is stored in the order of the sampled elapsed time.

  The sampling data record illustrated in FIG. 5A has fields such as “CPU number”, “PID”, “function identification information / instruction address”, and “number of consecutive times”. The “CPU number” and “PID” have been described with reference to the sampling data in FIG. In the “function identification information / instruction address” field, “function identification information” included in the function information table of the operation function specified by the second sampling processing unit 150 is stored. If the PID and instruction address are not included in the function information table, the “function identification information / instruction address” field contains “instruction position information that is collected by the second sampling processing unit 150 and that is address position information during execution of the instruction. Address "is stored. In the “continuous count” field, the continuous count of the sampling data counted by the continuous detection unit 160 is stored.

  In addition, in the “function identification information / instruction address” field, specific function identification information such as “A” or “0” newly given by the collected data storage unit 170 instead of “instruction address” is stored. It may be stored.

  In the example of FIG. 5A, the “CPU number” field stores the CPU number related to the execution of the program such as “0”. In the “PID” field, identification information of a process being executed such as “8” and “10” is stored. In the “function identification information / instruction address” field, function identification information of the operation function given by the function information table creation unit 140 such as “1” and “2” is stored. The “continuous number” field stores the continuous number of sampling data such as “10” and “8” counted by the continuous detection unit 160.

  The action function DB 220 stores action function information of action functions associated with OS functions, user functions, and the like. The operation function DB 220 has a record for each operation function. The action function information stored in the action function DB 220 includes a program name corresponding to the PID, a function name of the action function, a start address and an end address of the action function.

  Such operation function information may be set, for example, by an administrator of the sampling apparatus 100 via an operation input such as a keyboard, or may be set via a network connected to the sampling apparatus 100. .

As an example, in the case of an OS environment such as Linux, a list of OS function names and operation function start addresses and end addresses associated with the OS function names is stored in a file such as “/ proc / kallsyms”. An administrator of the sampling device 100 may access the above-described system file and acquire operation function information corresponding to each OS function.

For example, files such as “/ proc / PID / exe” and “/ proc / PID / stat” store the names of programs to be executed by each process. An administrator of the sampling apparatus 100 may access the above-described file and acquire a program name corresponding to each process.

Further, for example, by executing a predetermined command (for example, “objdump” command) on the execution file having the above program name, the operation function and address range (start address and end address) defined in the execution file are set. You can get it. The administrator of the sampling device 100 may acquire the operation function information defined by the user function by performing the above-described command operation.

  The administrator of the sampling apparatus 100 registers the action function information acquired by the above-described process in the action function DB 220 in association with the action function.

  FIG. 5B illustrates the function information DB 230. The function information table created by the function information table creation unit 140 is stored in the function information DB 230 of FIG. 5B. The function information DB 230 has a function information table for each target program.

  The function information table illustrated in FIG. 5B has fields such as “program name”, “function identification information”, “function name”, “start address”, and “end address”. The function information table illustrated in FIG. 5B has a record for each function identification information.

  In the example of FIG. 5B, the “program name” field stores a program name to be executed in the process corresponding to the PID. In the “function identification information” field, the function identification information of the operation function assigned by the function information table creation unit 140 is stored. The “function name” field stores the function name of the action function. The “start address” field stores the start address position of the motion function. The “end address” field stores the end address position of the action function.

In the example of FIG. 5B, program names such as “program1” and “program2” are stored in the “program name” field. In the “function identification information” field, function identification information of operation functions such as “1” and “2” is stored. The “function name” field stores function names of operation functions such as “Main” and “test”. The “start address” field stores the start address position of the operation function such as “100” and “3000”. The “end address” field stores the end address position of the operation function such as “250” and “3200”.

  For example, the second sampling processing unit 150 can identify the program name associated with “PID” by referring to the operation function DB 220 from “PID” included in the collected sampling data. Then, in the second sampling processing unit 150, the function information table illustrated in FIG. 5B is referred to from the position information of the “instruction address” included in the collected program data and the specified program name, and given to the operation function. Function identification information can be specified.

[Process flow]
Hereinafter, the sampling data recording control process of this embodiment will be described with reference to FIGS.

  FIG. 6 is an example of a flowchart relating to a function information table creation process performed by the sampling apparatus 100 according to the present embodiment. The flowchart of the example is a process mainly executed by the first sampling processing unit 110, the operation function specifying unit 120, the frequency calculating unit 130, and the function information table creating unit 140. For example, the sampling device 100 executes a function information table creation process by a computer program that is executably expanded in the main storage unit 102.

  In the flowchart illustrated in FIG. 6, the start of the function information table creation process can be exemplified by the activation of a computer program stored in the auxiliary storage unit 103, for example. The sampling apparatus 100 accesses a predetermined file and acquires OS function information in an OS environment in which the target program is executed (S1). The OS function information includes the OS function name and the start address and end address of the operation function associated with the OS function name.

  As described in the operation function DB 220, when the OS function information has already been acquired by the administrator of the sampling apparatus 100, the sampling apparatus 100 can omit the process of S1.

  In S2, the sampling apparatus 100 executes the target program and collects sampling data in a specific period shorter than the execution period of the target program. The sampling device 100 stores the collected sampling data in a predetermined area of the main storage unit 102, for example. Note that the processing of S2 has been described in the first sampling processing unit 110.

  In S3 to S5, the sampling apparatus 100 acquires the operation function information of the operation function from the PID and instruction address of the sampling data for the sampling data collected and collected in S2, and specifies the function name of the operation function. The sampling apparatus 100 repeats the processing of S3-S5 for the sampling data collected / collected in S2.

  In the processing of S3 to S5, the sampling apparatus 100 determines whether or not the instruction address of the sampling data collected in S2 is included in the address of the OS function (S3). The sampling apparatus 100 refers to the operation function DB 220 and determines whether or not the instruction address of the collected sampling data is included in the address range specified by the start address and end address of each OS function. When the instruction address of the sampling data is included in the address of the OS function (S3, Yes), the sampling device 100 specifies the function name of the operation function from the OS function information associated with the corresponding address range (S5). ). The sampling device 100 temporarily stores the function name of the identified operation function in a predetermined area of the main storage unit 102.

  On the other hand, when the instruction address of the sampling data is not included in the address of the OS function (S3, No), the sampling device 100 proceeds to S4.

  In S4, the sampling device 100 refers to the operation function DB 220 and specifies a program name associated with the PID of the collected sampling data. Then, the sampling device 100 identifies the function name of the action function associated with the action function DB 220 based on the address range of the action function associated with the identified program name and the instruction address of the collected sampling data. . The sampling device 100 temporarily stores the function name of the identified operation function in a predetermined area of the main storage unit 102.

  When the sampling device 100 completes specifying the function name of the operation function by the processing of S3-S5 for the sampling data collected and collected in S2, the sampling device 100 proceeds to S6.

  Note that the processing of S3 to S5 executed by the sampling device 100 is an example of an acquisition step and an acquisition procedure for acquiring function identification information. In addition, the sampling apparatus 100 performs the processing of S3 to S5 as an example of an acquisition unit that acquires function identification information.

In S6, the sampling apparatus 100 counts the number of samples of each action function specified in S3-S5, and calculates a sample total value for each action function. Then, the sampling apparatus 100 performs a mutual comparison of the sample aggregate values counted for each operation function, and arranges the operation functions in the order of, for example, the sample aggregate values in descending order. Each action function arranged in order of increasing sample aggregate value is stored in a predetermined area of the main storage unit 102 together with the sample aggregate value.

  Further, the processing of S1-S6 executed by the sampling device 100 acquires the function identification information and the number of times of sampling of the function identification information, and acquires the function identification information and the number of times of sampling of the function identification information. It is an example of a procedure. In addition, the sampling apparatus 100 performs the processing of S1 to S6 as an example of a unit that acquires the function identification information and the number of times the function identification information is sampled.

  In S7, the sampling device 100 selects an operation function based on the sample total value counted in S6, adds function identification information to the selected operation function, and creates a function information table. The created function information table is exemplified in FIG. 5B, for example. For example, the sampling apparatus 100 can select a higher-order operation function arranged in the descending order of sample count values in S6. It should be noted that the number of higher-order operation functions can be selected arbitrarily. A higher-order operation function is an example of an operation function that satisfies a predetermined criterion.

  Note that the processing of S7 executed by the sampling apparatus 100 is an example of a step of creating a function information table and a procedure of creating a function information table. In addition, the sampling device 100 executes the process of S7 as an example of a means for creating a function information table.

  For example, the sampling apparatus 100 can set a predetermined threshold value, and can set an operation function whose sample total value exceeds the predetermined threshold value as a higher-level operation function. Further, for example, the sampling apparatus 100 may select a higher-order operation function that is arranged in order of increasing sample count value in S6 along a preset number (for example, 20). For example, the sampling apparatus 100 may select a higher-order operation function that occupies a certain ratio (for example, 50% of the entire sample) among the operation functions totaled in S6.

  The sampling apparatus 100 only needs to be able to identify an operation function having a large sample count value counted in S6. This is because an operation function having a large sample aggregate value can be identified as an operation function having a high sampling frequency during the execution period of the target program.

  The sampling device 100 assigns function identification information that uniquely identifies the motion function to the selected higher-order motion function. Then, the sampling apparatus 100 creates a function information table by associating the assigned function identification information with the action function name, the program name corresponding to the PID, the start address and the end address position of the action function. The created function information table is stored in the function information DB 230.

  With the above processing, the sampling apparatus 100 can identify an operation function with a high sampling frequency during the execution period of the target program from the sampling data collected and collected in S2. Each identified operation function is given function identification information. The assigned function identification information and the function information unique to the action function are associated with each other and stored in the function information DB 230 as a function information table. The sampling device 100 can shorten the processing time for detecting the operation function by referring to the function information table stored in the function information DB 230 during the execution period of the target program.

Next, sampling data recording processing according to the present embodiment will be described with reference to the flowchart of FIG. FIG. 7 is an example of a flowchart relating to sampling data recording processing by the sampling device 100 of the present embodiment. The flowchart in the example is a process mainly executed by the second sampling processing unit 150, the continuous detection unit 160, and the collected data storage unit 170. The sampling device 100 executes sampling data recording processing by, for example, a computer program developed in the main storage unit 102 so as to be executable.

  In the flowchart illustrated in FIG. 7, the start of the sampling data recording process (“sampling start” in the example) can be exemplified by the activation of a computer program stored in the auxiliary storage unit 103. In the sampling device 100, each functional unit such as the second sampling processing unit 150, the continuous detection unit 160, and the collected data storage unit 170 is executed in parallel with the execution of the target program by the activation of the computer program.

  In the illustrated flowchart, the processes of S8 to S17 are repeatedly executed every time sampling data is collected at a predetermined sampling period during the execution period of the target program. That is, the sampling data recording process by the sampling apparatus 100 according to the present embodiment ends with the end of the execution of the operation of the target program.

  In the process of S8-S17, the sampling apparatus 100 collects information related to the operation of the target program as sampling data at a predetermined sampling period (S8). Note that the sampling data collected in S8 is collected, for example, in the data format illustrated in FIG. As illustrated in FIG. 2, the information related to the operation of the target program includes “CPU number”, “PID”, and “instruction address”. The collected sampling data is temporarily stored in a predetermined area of the main storage unit 102.

  In S9, the sampling device 100 refers to the operation function DB 220 from the PID included in the sampling data collected in S8, and identifies the program name associated with the PID. Then, the sampling device 100 ends from the instruction address included in the sampling data collected in S8 and the specified program name, from the start address of the operation function defined in the function information table in which the sampling data is stored in the function information DB 230. It is determined whether it is within the address range (S9-S10).

  In S10, the sampling apparatus 100 refers to the function information table corresponding to the target program stored in the function information DB 230, and determines whether the sampling data is an operation function included in the function information table.

  For example, the sampling device 100 searches the function information table based on the program name specified in S9, and obtains a start address and an end address associated with the program name. Then, the sampling apparatus 100 may determine whether the address in the program included in the sampling data is included in the position range that can be specified by the start address and the end address. The sampling apparatus 100 may determine that the sampling data corresponds to the operation function included in the function information table when the program name and the instruction address are included in the function information table.

  The sampling apparatus 100 may store the function information table corresponding to the target program stored in the function information DB 230 in a predetermined area of the main storage unit 102 together with the start of sampling.

  When the instruction address specified from the sampling data corresponds to the operation function included in the function information table (S10, Yes), the sampling device 100 acquires function identification information corresponding to the operation function from the function information table. Then, the sampling device 100 stores the acquired function identification information in a predetermined area of the main storage unit 102 (S12).

On the other hand, when the instruction address specified from the sampling data does not correspond to the operation function included in the function information table (S10, No), the sampling apparatus 100 uses the instruction address specified in S9 as the predetermined address in the main storage unit 102. Store in the area (S11).

  Next, in S13, the sampling apparatus 100 compares the sampling data being processed with the sampling data collected in the immediately preceding sampling period, and determines whether the sampling data being processed is continuous.

  Here, continuous sampling data means (1) when sampling data of the same operation function is detected continuously, (2) sampling data of the same PID and instruction address are detected continuously Say the case.

  For example, the sampling apparatus 100 compares the CPU number and operation function of the sampling data being processed with the CPU number and operation function of the sampling data collected in the immediately preceding sampling cycle. The sampling apparatus 100 can determine whether the sampling data is continuous in time series order by determining the coincidence of both CPU numbers and operation functions.

  Further, for example, the sampling apparatus 100 compares the PID and instruction address of the sampling data being processed with the PID and instruction address of the sampling data collected in the immediately preceding sampling cycle. The sampling apparatus 100 can determine whether the sampling data is continuous in chronological order by determining the coincidence of both PIDs and instruction addresses.

  When the operation function of the sampling data being processed matches the operation function of the immediately preceding sampling data (S13, Yes), the sampling device 100 sets, for example, a predetermined area of the main storage unit 102. The count value of the continuous detection counter is incremented by “1”. Here, as a continuous detection counter set in a predetermined area of the main storage unit 102, the sampling device 100 holds, for example, the records of each row shown in the table of FIG. 5A in the predetermined area of the main storage unit 102. Just keep it. Then, the “continuous count” field in each row shown in FIG. 5A may be used as a continuous detection counter.

  Then, the sampling device 100 stores the count value in a predetermined area (S16). Note that the sampling device 100 continuously increments the count value of the continuous detection counter by “1” every time it detects the sampling data that coincides continuously, so that the number of times the operation function (or PID and instruction address) of the sampling data continues. Can be counted.

  On the other hand, when the operation function of the sampling data being processed does not match the operation function of the immediately preceding sampling data (No at S13), the sampling device 100 converts the sampling data being processed to new sampling data. Is stored in a predetermined area of the main storage unit 102 (S14). For example, if the sampling apparatus 100 secures a predetermined area of the main storage unit 102 in the same format as the record of each row shown in the table of FIG. 5A and sets the initial value to 1 in the “continuous number of times” field. Good.

  Note that the processing in S13 to S15 is the same even when it is determined that the PID and the instruction address match. If the PID and instruction address of the sampling data being processed match the PID and instruction address of the immediately preceding sampling data (S13, Yes), the sampling apparatus 100 may, for example, store the predetermined data in the main storage unit 102. The count value of the continuous detection counter set in this area is incremented by “1” and the count value is stored in a predetermined area. On the other hand, when the PID and instruction address of the sampling data being processed do not match the PID and instruction address of the immediately preceding sampling data, the sampling apparatus 100 sets the sampling data being processed as new sampling data. What is necessary is just to store in the predetermined area | region of the main memory part 102. FIG.

  Here, the process of S13 executed by the sampling apparatus 100 is an example of a detection step and a detection procedure. Further, the CPU 101 or the like of the sampling apparatus 100 executes the process of S13 as an example of a detection unit. Moreover, the process of S15 performed by the sampling apparatus 100 is an example of a counting step and a counting procedure. In addition, the CPU 101 and the like of the sampling device 100 execute the process of S15 as an example of a counting unit. Furthermore, the processes of S14 and S16 executed by the sampling apparatus 100 are an example of a recording step and a recording procedure. In addition, the CPU 101 and the like of the sampling device 100 execute the processes of S14 and S16 as an example of a recording unit.

  Note that if the operation function cannot be identified in the determination of S10, the sampling device 100 may add the attached identification information to the sampling information instead of the instruction address in S11. The attached identification information is information indicating that the function could not be specified from the address of the sampling information. As the attached identification information, the sampling apparatus 100 may create information for collecting instruction addresses such as “A” and “0” to create a sampling data record. The attached identification information for collecting the instruction addresses may be identified from the function identification information specified in S12. In this case, the process of S11 executed by the sampling device 100 is an example of an attached identification information adding step and an attached identification information adding procedure. In addition, the CPU 101 and the like of the sampling device 100 execute the process of S11 as an example of an attached identification information adding unit.

  In S17, the sampling apparatus 100 determines the end of the operation execution of the target program. If the operation execution is not ended, the sampling apparatus 100 proceeds to S8 (S17, No), and repeats the processing of S8-S17. On the other hand, the sampling device 100 ends the sampling data recording process when the operation execution of the target program is ended (S17, Yes).

  As described above, in the sampling data recording process of the present embodiment, when the operation function of the sampling data is consecutive in the time-series order in which the sampling data is sampled, the number of consecutive times is stored instead of the sampling data. The amount of sampling data at the time of sampling exemplified by the equation Also, in the sampling data recording process of the present embodiment, even when the PID and instruction address of the sampling data are consecutive in the time series in which they are sampled, the number of consecutive times is stored instead of the sampling data. Can be reduced. For example, it is possible to reduce the amount of data collected and accumulated by one sampling for the target program. In particular, when the target program includes an idle function that does not process anything during the program execution period or a lock-related function that waits for the occurrence of a predetermined event, the reduction of the data amount is effective.

  In addition, for example, even if the computer that executes the target program is a computer such as a multiprocessor having a plurality of processors or a multicore processor in which a plurality of cores are incorporated in a single CPU package, long-term sampling data can be collected. I can expect.

  In addition, for example, in a technical field such as HPC (High-performance computing), long-time sampling data can be expected to be collected even for a target program such as a mathematical calculation program executed by a supercomputer or the like. .

  Further, in the sampling data recording process of the present embodiment, an operation function with a relatively high sampling frequency can be referred to as a function information table, so that the processing time for detecting the operation function can be shortened. In the sampling data recording process of the present embodiment, the operation functions included in the function information table are to be collected and collected, and the operation functions not included in the function information table are excluded from the collection and collection. If this is the case, further reduction of data volume can be expected.

  Furthermore, in the sampling data recording process of the present embodiment, when collecting an operation function that is not included in the function information table, function identification information that identifies the instruction address instead of the instruction address is newly added, Sampling data records can be created. For this reason, as exemplified in the equation (1), one record data size can be reduced as compared with the case where the instruction address is stored.

<Modification>
As a modification of the sampling data recording process of the present embodiment, the count values may be collected for each basic block divided by the conditional branch of the target program. For example, the sampling apparatus 100 may divide the target program into basic blocks corresponding to the branch destination of the branch instruction. The sampling apparatus 100 may create a classification table including basic block identification information, the address of the first instruction of the basic block, and the address of the last instruction of the basic block.

  Then, the sampling device 100 determines whether the address of the collected sampling data is between the head address and the end address of the basic block during the execution period of the target program, and the basic corresponding to the sampling data What is necessary is just to give the identification information of a block. The procedure for assigning basic block identification information is the same as the procedure for assigning function identification information in the above embodiment.

  The sampling apparatus 100 can collect the count value of the sampling data finer than the function by recording the count value of the identification information of the basic block.

<Computer-readable recording medium>
A program for causing a computer or other machine or device (hereinafter, a computer or the like) to realize any of the above functions can be recorded on a recording medium that can be read by the computer or the like. Then, the function can be provided by causing the computer or the like to read and execute the program of the recording medium.

  Here, a computer-readable recording medium is a recording medium that stores information such as data and programs by electrical, magnetic, optical, mechanical, or chemical action and can be read from a computer or the like. Say. Examples of such a recording medium that can be removed from a computer or the like include a flexible disk, a magneto-optical disk, a CD-ROM, a CD-R / W, a DVD, a Blu-ray disk, a DAT, an 8 mm tape, a flash memory, and the like. There are cards. Moreover, there are a hard disk, a ROM, and the like as a recording medium fixed to a computer or the like.

<Others>
The above embodiment further includes an aspect called the following supplementary note. The components included in the following supplementary notes can be combined with the constituents included in the other supplementary notes.

(Appendix 1)
The sampling information related to the operation of the target program collected at a predetermined sampling period during the execution period of the target program to be sampled is compared between adjacent sampling information in chronological order, and has common elements A detection step for detecting that the sampling information is continuous in chronological order;
A counting step of counting the number of consecutive samplings of the sampling information detected to be continuous in chronological order in the detection step;
A recording step in which the sampling information detected to be continuous in chronological order in the detection step and the continuous sampling count counted in the counting step are recorded in association with each other;
Sampling program to execute.

(Appendix 2)
Based on a function information table including identification information of a function in the target program and an address range in which the function is arranged in the target program, the sampling is performed from an address in the target program included in the sampling information. And further executing an acquisition step of acquiring identification information of the function corresponding to the information,
The sampling program according to appendix 1, wherein in the recording step, the obtained function identification information and the number of times the function identification information is sampled are recorded.

(Appendix 3)
The target program is executed in a specific period, information related to the operation of the target program is collected at a predetermined sampling period during the specific period, and function identification information and function in the target program are collected from the collected sampling information Obtaining a sampling number of identification information of
The sampling program according to appendix 2, for further executing the step of creating the function information table for a function whose sampling number of identification information of the function satisfies a predetermined criterion.

(Appendix 4)
The obtaining step includes an attached identification information giving step of giving attached identification information of sampling information in place of the function identification information when the identification information of the function corresponding to the sampling information cannot be obtained,
The recording step associates the identification information of the operation function acquired in the acquisition step or the auxiliary identification information given in the auxiliary identification information giving step with the continuous sampling count counted in the counting step. The sampling program according to appendix 2 or 3, which is recorded.

(Appendix 5)
Computer
The sampling information related to the operation of the target program collected at a predetermined sampling period during the execution period of the target program to be sampled is compared between adjacent sampling information in chronological order, and has common elements A detection procedure for detecting that sampling information is continuous in chronological order;
A counting procedure for counting the number of consecutive samplings of the sampling information detected to be continuous in chronological order in the detection procedure;
A recording procedure for recording the sampling information detected to be continuous in chronological order in the detection procedure and the number of consecutive samplings counted in the counting procedure;
Perform sampling method.

(Appendix 6)
Based on a function information table including identification information of a function in the target program and an address range in which the function is arranged in the target program, the sampling is performed from an address in the target program included in the sampling information. Further executing an acquisition procedure for acquiring the identification information of the function corresponding to the information,
6. The sampling method according to appendix 5, wherein in the recording procedure, the acquired function identification information and the number of sampling times of the function identification information are recorded.

(Appendix 7)
The target program is executed in a specific period, information related to the operation of the target program is collected at a predetermined sampling period during the specific period, and function identification information and function in the target program are collected from the collected sampling information To obtain the sampling number of identification information of
The sampling method according to appendix 6, further executing a procedure of creating the function information table for a function whose sampling number of identification information of the function satisfies a predetermined criterion.

(Appendix 8)
The acquisition procedure includes an attached identification information providing procedure for adding attached identification information of sampling information instead of the function identification information when the identification information of the function corresponding to the sampling information cannot be acquired,
The recording procedure associates the identification information of the operation function acquired in the acquisition procedure or the attached identification information given in the attached identification information addition procedure with the continuous sampling count counted in the counting procedure. The sampling method according to appendix 6 or 7, which is recorded.

(Appendix 9)
The sampling information related to the operation of the target program collected at a predetermined sampling period during the execution period of the target program to be sampled is compared between adjacent sampling information in chronological order, and has common elements Detecting means for detecting that the sampling information is continuous in chronological order;
Counting means for counting the number of consecutive samplings of the sampling information detected to be continuous in chronological order by the detection means;
Recording means for recording the sampling information detected to be continuous in time series by the detection means and the number of consecutive samplings counted by the counting means,
An information processing apparatus comprising:

(Appendix 10)
Based on a function information table including identification information of a function in the target program and an address range in which the function is arranged in the target program, the sampling is performed from an address in the target program included in the sampling information. An acquisition means for acquiring identification information of a function corresponding to the information;
The information processing apparatus according to appendix 9, wherein the recording unit records the acquired function identification information and the number of times the function identification information is sampled.

(Appendix 11)
The target program is executed in a specific period, information related to the operation of the target program is collected at a predetermined sampling period during the specific period, and function identification information and function in the target program are collected from the collected sampling information Means for obtaining the number of times the identification information is sampled,
The information processing apparatus according to appendix 10, further comprising means for creating the function information table for a function whose sampling number of identification information of the function satisfies a predetermined criterion.

(Appendix 12)
The acquisition means includes attached identification information giving means for giving attached identification information of sampling information in place of the function identification information when the function identification information corresponding to the sampling information cannot be obtained,
The recording means associates the identification information of the operation function acquired by the acquisition means or the attached identification information given by the attached identification information giving means with the number of consecutive samplings counted by the counting means. The information processing apparatus according to appendix 10 or 11, which is recorded.

100 Sampling device 101 CPU
102 main storage unit 103 auxiliary storage unit 104 communication unit 110 first sampling processing unit 120 operation function specifying unit 130 frequency calculation unit 140 function information table creation unit 150 second sampling processing unit 160 continuous detection unit 170 collected data storage unit 200 storage device 210 Sampling DB
220 Action function DB
230 Function information DB

Claims (4)

On the computer,
Sampling information having common elements by comparing the sampling information related to the operation of the target program collected at a predetermined sampling period during the target program execution period to be sampled, by comparing adjacent sampling information in time series order A detecting step for detecting that is continuous in chronological order;
A counting step of counting the number of consecutive samplings of the sampling information detected to be continuous in chronological order in the detection step;
Based on a function information table including identification information of a function in the target program and an address range in which the function is arranged in the target program, the sampling is performed from an address in the target program included in the sampling information. An acquisition step of acquiring identification information of a function corresponding to the information;
The identification information of the function corresponding to the sampling information detected to be continuous in chronological order in the detection step and the number of consecutive samplings counted in the counting step are collected in association with each other, from the collected sampling information, Obtaining the identification information of the function in the target program and the sampling count of the identification information of the function;
Creating a function information table for a function whose number of samplings of identification information of the function satisfies a predetermined criterion ;
Sampling program to execute. The obtaining step, if it can not obtain the identification information of the function corresponding to the sampling information comprises accessory identification information imparting step of imparting the supplied identification information of the sampling information in place of the identification information of the function,
The identification information of the function corresponding to the sampling information detected to be continuous in chronological order in the detection step , or the attached identification information given in the attached identification information giving step, and the number of consecutive samplings counted in the counting step The sampling program according to claim 1 , which is recorded in association with each other . Computer
Collected at a predetermined sampling period during the target program execution period to be sampled
For the sampling information related to the operation of the target program, a detection step of comparing between adjacent sampling information in time series order and detecting that sampling information having common elements is continuous in time series order; and
A counting step of counting the number of consecutive samplings of the sampling information detected to be continuous in chronological order in the detection step;
Based on a function information table including identification information of a function in the target program and an address range in which the function is arranged in the target program, the sampling is performed from an address in the target program included in the sampling information. An acquisition step of acquiring identification information of a function corresponding to the information;
The identification information of the function corresponding to the sampling information detected to be continuous in chronological order in the detection step and the number of consecutive samplings counted in the counting step are collected in association with each other, from the collected sampling information, Obtaining the identification information of the function in the target program and the sampling count of the identification information of the function;
Creating a function information table for a function whose number of samplings of identification information of the function satisfies a predetermined criterion ;
Sampling method for the execution. Sampling information related to the operation of the target program collected at a predetermined sampling period during the execution period of the target program to be sampled is compared between adjacent sampling information in time series order, and has a common element Detecting means for detecting that information is continuous in time series; and
Counting means for counting the number of consecutive samplings of the sampling information detected to be continuous in chronological order by the detection means;
Based on a function information table including identification information of a function in the target program and an address range in which the function is arranged in the target program, the sampling is performed from an address in the target program included in the sampling information. Obtaining means for obtaining identification information of a function corresponding to the information;
The identification information of the function corresponding to the sampling information detected to be continuous in chronological order by the detection means and the number of consecutive samplings counted by the counting means are collected in association with each other, from the collected sampling information, Means for obtaining the identification information of the function in the target program and the number of sampling times of the identification information of the function;
Means for creating the function information table for a function whose number of samplings of identification information of the function satisfies a predetermined criterion;
An information processing apparatus comprising:

Images