JP3318121B2 - Virtual computer system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a virtual computer system for tracing instructions in each virtual computer under a virtual computer control system and collecting information unique to the system on each virtual computer.

[0002]

2. Description of the Related Art FIG. 9 is a diagram for explaining a mechanism of instruction tracing in a virtual machine system. Instruction tracing in the virtual machine system is realized by using a mechanism called program event recording (hereinafter referred to as PER interrupt), which is one of the program monitoring functions.

[0003] The PER interrupt includes an instruction reading event from the main storage device. If an address range of the main storage device is designated when setting the PER interrupt in hardware, the program is executed after the instruction in the range is executed. An interrupt occurs. Instruction tracing is performed using the above mechanism as shown in FIG. Note that the PER interrupt is not controlled by the virtual machine control system itself, but the virtual machine control system requests hardware that realizes the virtual machine system to control the PER interrupt.

In FIG. 9, instruction tracing is performed as follows. Normally, the virtual machine control system AVM assigns the CPU to the virtual machine VM1 which is one of its subordinate virtual machines for a certain period of time, and the system of the virtual machine VM1 operates. Note that while the virtual machine VM1 is operating, the virtual machine control system AVM does not intervene, and the control returns to the virtual machine control system AVM due to the interruption of the passage of time. When there is a trace start instruction, a PER interrupt is set, and the CPU is assigned to the virtual machine VM1. Thereafter, each time the instruction of the virtual machine VM1 is executed, an interrupt occurs, and the control returns to the virtual machine control system AVM.

At this time, the virtual machine control system AVM outputs the instruction that caused the interrupt and other information as trace information. When there is a trace end instruction, the virtual machine control system AVM releases the PER interrupt and returns to the normal operation. FIG. 10 is a diagram showing a conventional instruction tracing mechanism in the above-described virtual machine system. The conventional instruction tracing will be described with reference to FIG. The numbers in parentheses below correspond to the numbers in parenthesis in FIG. (1) Preparation for Trace Start In the spool of the virtual machine control system, all the parts used for the own virtual machine VM1, which is the target of the instruction trace, are deleted, and an unused area is generated. In addition,
The spool is an area for storing a list output from each virtual machine provided on the virtual machine control system, and is also used as an area for storing trace data. (2) Instruct the virtual machine control system to start tracing. (3) The virtual machine control system AVM outputs all instructions on the virtual machine VM1 for which the trace instruction has been issued to the spool as trace data. (4) After tracing, the local virtual machine VM1 in the spool
Only the output of is backed up to a magnetic tape or the like.

When tracing is performed again, the above operation is repeated. (5) Subsequently, when tracing is performed for the next virtual machine VM2, the traceable amount is the data amount that can be stored in the remaining unused area of the spool.
The trace stops when the spool becomes full.

[0007]

The instruction tracing mechanism in the above-mentioned conventional virtual computer system has the following problems. (A) There is only one spool on the virtual machine system, it is shared by all virtual machines, and there is no free space unless the output of each virtual machine is deleted. This is the same for the output of the own virtual machine.

For this reason, when the spool becomes full, it is impossible to output a list including the instruction trace of each virtual machine thereafter. (B) Since the spool is shared by all the virtual machines as in (a) above, the traceable amount is up to the amount of data that falls into the currently unused area of the spool. (C) When tracing a plurality of virtual machines, the trace data is output to the spool unused area of each virtual machine. Therefore, if two virtual machines start tracing at the same time, only half of the unused area at the start of tracing can store trace data, and if n virtual machines start tracing at the same time, Only 1 / n of the unused area can store trace data. (D) When performing repeated tracing, it is necessary to suck the previous trace result from the spool onto a magnetic tape or the like and delete it.

In particular, if instruction tracing is continuously performed, it becomes very burdensome to determine where the trace data is from the first time and from where to the second time. As described above, the instruction tracing mechanism in the conventional virtual machine system uses one spool shared by all virtual machines as a buffer for the measurement result, and stops the measurement when the spool becomes full as described above. Need to output data. Therefore, there is a problem that the number of steps that can be measured is limited depending on the size of the spool and the free space. In addition, since the spool is not assigned to each virtual machine, when tracing instructions of a plurality of virtual machines, there is a problem that a measurement result needs to be separated for each virtual machine.

In addition to the above problems, the conventional virtual computer system has the following problems. (1) In the conventional virtual machine control system, since the space switch command of the virtual machine was not considered,
Space information could not be collected, and the result of instruction tracing by the virtual machine control system could not be identified for each space, and the number of running steps could not be grasped for each space. (2) In the conventional virtual machine control system, there is no means for obtaining information such as the creation of a space during measurement or the loading of modules, and the instruction tracing mechanism of the virtual machine control system as described above. In the measurement of the number of traveling steps using the method, it is not possible to collect dynamically loaded module information (address size), and it is impossible to grasp the exact number of steps for each module.

Conventionally, as a method of collecting module information dynamically loaded during measurement, there has been a method of sampling module information of the entire space at fixed time intervals. The running time of the loading module could not be accurately grasped. In addition, the overhead of module information collection is large, and since the module information is collected on the virtual machine to be measured, the number of steps of the module information collection processing itself is included in the instruction trace data. There was a problem that the correction processing could not be completed. (3) In the conventional virtual machine control system, there is no mechanism for outputting information unique to the trace data by a specific hardware instruction even if it is desired to output information unique to the trace data.

For this reason, information that dynamically changes during measurement is collected on the virtual machine to be measured, and even if collected, instructions that the collection process itself runs are traced, and a large amount of useless instruction trace data is collected. There was a problem of being collected, and in fact, he had given up collecting. (4) In an OS such as UNIX (product name) in which a hardware instruction for switching a virtual space cannot be specified (for example, space switching is performed by a combination of a plurality of hardware instructions), a dynamic step measurement method Accumulates only program counter information (information on areas where hardware controls execution of hardware instructions, such as instruction addresses, interrupt masks, operation modes, etc.). Was a method of distributing dynamic steps to At this time, since the information that the space was switched and the information about which memory map to use were not collected, the dynamic steps could not be assigned to each space.

However, even in the prior art, whether the OS control program or the user program is running can be determined from the program counter information, so that the dynamic steps are counted only for the OS control program. It was possible. (5) By the way, in an OS such as UNIX (product name) for obtaining a memory map from a symbol table of an executable file, in order to obtain a memory map,
Information on which file to use must be collected as space identification information. Further, in the UNIX system, a file is identified by a name corresponding to a file generally called a path name.

When storing the information of the memory map change, it is necessary to store the location of the memory map change. However, in the above-mentioned OS, the path length is not fixed, so that the data to be stored is There was a risk of increasing the amount. SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems of the prior art, and a first object of the present invention is to make an instruction tracing mechanism independent for each virtual machine to be measured, and to measure a plurality of virtual machines. It is to provide a virtual computer system which enables the above.

A second object of the present invention is to provide a mechanism for collecting space information in a control register when executing a hardware instruction in which space switching occurs, and to provide a hardware instruction executed after the space switching for each space. It is to provide a virtual computer system that can be analyzed. A third object of the present invention is to provide a virtual computer system capable of collecting information on dynamically loaded modules and grasping the number of running steps for each spatial module.

A fourth object of the present invention is to provide a virtual computer system capable of outputting information peculiar to the system as instruction trace data with minimum overhead. A fifth object of the present invention is to provide a virtual machine to be measured having an OS that cannot specify a hardware instruction for switching a virtual space.
An object of the present invention is to provide a virtual computer system capable of grasping the number of steps.

A sixth object of the present invention is to provide a virtual machine system having a virtual machine to be measured having an OS in which a variable-length identifier is used as information for identifying a file. An object of the present invention is to provide a virtual computer system that can prevent an increase in stored data by converting the data into a fixed-length identifier indicating the physical location of the virtual computer.

[0018]

FIG. 1 is a block diagram showing the principle of the present invention. In the figure, AVM is a virtual machine control system, V1 to V3 are virtual machines to be measured, and are subject to instruction tracing. The SVM is a server virtual machine that collects trace data and system-specific information. The server virtual machine SVM is a virtual machine to be measured (for performing a trace.
It receives the instruction trace information of the target virtual machine) and outputs it to an external medium.

J1 to J3 are server jobs of the server virtual machine SVM. The server jobs J1 to J3 are started for each of the plurality of virtual machines to be measured, and the trace buffers B1,
The trace data and the system-specific information are output to B2. The AS is an analysis system that performs analysis based on trace data and system-specific information. The analysis system can be provided for each of the server jobs J1 to J3.

[0020] To solve the above problem, as shown in Figure 1, the invention of claim 1 of the present invention, in the virtual computer system for performing the instruction tracing of a plurality of virtual machines under the virtual machine control system, the virtual machine Control system
The virtual instrument to be traced under the stem
The instruction trace for each computer is executed by the virtual machine control system.
And collect the results of this
A server virtual computer for outputting to a trace buffer corresponding to a virtual computer is provided .

According to a second aspect of the present invention, in the first aspect of the present invention, the trace buffer is located before the trace target.
A first trace buffer for each virtual machine and
A second trace buffer, the output of said output means.
Of the first buffer or the second buffer
Output to the output means.
The first buffer or the second
Trace data stored in the buffer to an external medium
And second output means for outputting. According to a third aspect of the present invention, in the first or second aspect, the server virtual machine collects space identification information in a control register when executing a hardware instruction that causes a space switch, and traces the trace. A mechanism for outputting data is provided.

The invention according to claim 4 of the present invention relates to claims 1 and 2.
Alternatively, in the invention according to claim 3, an object to be traced
A first exit function which is called when the space is created and registers a loading completion exit in the created space, and stores module information when module loading is completed, and And a second exit function for outputting as trace data by using an interface of the virtual machine control system by placing the information in a register, and collecting dynamically loaded module information.

The invention of claim 5 of the present invention is directed to claim 1,
In the second, third or fourth aspect of the present invention, tracing is performed.
A measurement setup job is provided in the virtual machine to be processed, module information of a space where the measurement setup job exists before the start of tracing is set in a management table, and a loading completion exit is registered. is there. The invention of claim 6 of the present invention is the invention of claim 1, 2, 3, 4 or claim 5, a mechanism that outputs register information at the time of hardware instruction execution as trace data, and subjected to the trace Become
On the virtual machine, when necessary information specific to the system is stored in a register, the hardware instruction is executed to collect necessary system specific information.

According to a seventh aspect of the present invention, a first aspect is provided.
In the second, third, fourth, fifth or sixth aspect of the present invention, a hardware instruction for switching the virtual space cannot be specified.
In the virtual computer system having a virtual machine having a S, when the hardware instruction is executed by the virtual machine in question to perform the tracing, the mechanism for outputting the identification information of the switched space as trace data is provided, A mechanism is provided for ascertaining the number of dynamic steps for each virtual space based on the space identification information output by executing the above hardware instruction at the time of space switching.

The invention according to claim 8 of the present invention is directed to claim 1,
In the invention according to claim 2, when measuring a virtual machine having an OS using a variable-length identifier as information for identifying a file, the variable-length identifier is used. Instead, a fixed-length identifier indicating the physical location of a file is used, and the fixed-length identifier is stored to secure information for creating a memory map.

[0026]

[Action]

(A) Function of Server Virtual Machine In the figure, prior to instruction tracing, the user executes a job J for the virtual machine to be measured on the server virtual machine SVM.
1 to J3 are activated. If the virtual machine to be measured is a multi-CPU, by instructing the CPU that performs instruction tracing to the server jobs J1 to J3,
Tracing for each CPU becomes possible.

The buffers B1 to B3 for storing the instruction trace results are prepared for each virtual machine to be measured by each of the server jobs J1 to J3, so that mutual influence is exerted even when other virtual machines perform simultaneous measurement. There is no restriction on the amount of measurement data. When performing instruction tracing, the server job J
1 to J3 receive the trace information of the designated virtual machines VM1 to VM3 from the virtual machine control system AVM via the buffer, and store them in the external medium D. Further, two buffers B1 to B3 are prepared as shown in FIG. 1, and are used while switching between writing and reading.

When the buffers B1 to B3 are full, the virtual machine control system AVM waits until a free buffer is notified.
Stops the dispatch of the virtual machines to be measured VM1 to VM3 (since the execution right is not given to the virtual machines to be measured), no data loss occurs. As described above, instead of outputting the contents of the buffers B1 to B3 to the external medium D, it is possible to directly analyze the contents using the server jobs J1 to J3. However, depending on the contents to be analyzed, the server job is required for the analysis. Since time may be an overhead of instruction tracing, it is necessary to properly use the analysis contents such as simple analysis or detailed analysis. (B) Function of virtual machine to be measured Virtual machines to be measured VM1 to V after instruction trace start instruction
Hardware instructions executed in M3 are traced by the virtual machine control system AVM, and the server jobs J1 to J
3 is notified to the server jobs J1 to J3 via the buffer prepared.

At this time, an interface for notifying the server job of system-specific information as instruction trace data in the virtual machine control system AVM is constructed, and system-specific information of the measured virtual machines VM1 to VM3 is output as trace data. Trace data includes hardware instruction information, interrupt information, space identification information,
There are dynamic module information, area acquisition / return information, job information, and the like.

Since the virtual machine to be measured is traced by instructions, it is necessary to realize an interface for outputting system-specific information and a process for collecting the system-specific information itself in the minimum running steps. In the present invention, a mechanism for outputting system-specific information and a mechanism for collecting system-specific information for this purpose have been constructed. In addition, on the measured virtual machine,
The collection of auxiliary data for detailed analysis from the measured data is also performed during or after the measurement. (C) Analysis Function The analysis function AS counts the number of running steps for each space, module and section based on the trace data collected by the server virtual machine and the analysis auxiliary data collected by the measured virtual machine. The system in which the analysis function operates may be any system.

[0031]

【Example】

A. Instruction tracing mechanism and collection mechanism of a plurality of virtual machines Execution processing in each virtual machine under the virtual machine control system is independent between the virtual machines. Therefore, in the present embodiment, a virtual machine control system and a server virtual machine that performs a specific job in response to a request from a virtual machine under the virtual machine control system are provided, and a plurality of virtual machines are measured by one server virtual machine as shown below. Was made possible. Thus, the instruction tracing mechanism can be an independent tracing mechanism for each virtual machine to be measured.

FIG. 2 is a diagram showing this embodiment. In FIG. 2, AVM is a virtual machine control system for tracing instructions of a designated virtual machine, SVM is a server virtual machine, and server virtual machine SVM is It receives instruction trace information of the measurement virtual machine and stores it in an external medium described later. J1 to J3 are virtual machine control systems A
This is a server job that has an interface with the VM and performs processing to be described later. The server jobs J1 to J3 are provided corresponding to the respective virtual machines to be measured, and are activated for each virtual machine to be measured.

VM1 to VM3 are instruction traces, and
Virtual computers B1 to B3 which are targets of collecting system-specific information to be described later are trace buffers provided corresponding to the virtual machines VM1 to VM3. The trace buffer has two buffers, a buffer 1 and a buffer 2. .
D is an external storage medium made of DASD or magnetic tape.

In this embodiment, as shown in FIG. 3, each of the virtual machines VM1 to VM3 is placed on the server virtual machine SVM.
The server jobs J1, J2, and J3 corresponding to the above are provided, and an environment in which these server jobs trace an instruction of each virtual machine to be measured is constructed to collect information. As a result, a plurality of virtual machines can be measured by one server virtual machine SVM.

Next, the instruction trace processing by the server job will be described. Note that the numbers in parentheses (1), (2),... In FIG. (1) The server job J1 is a virtual machine VM to be measured.
1 is notified to the virtual machine control system AVM. (2) The virtual machine control system AVM creates a measurement environment for the virtual machine VM1 to be measured. (3) The measured virtual machine VM1 instructs the start of tracing. (4) The virtual machine control system AVM uses the server job J1.
Is notified that there is a trace start instruction. (5) The server job J1 secures a trace buffer of the virtual machine VM1, and notifies the virtual machine control system AVM of the buffer address and size. (6) The virtual machine VM1 starts instruction tracing, and the trace data is stored in the trace buffer B1 provided for each virtual machine to be measured via the virtual machine control system AVM. (7) Thereafter, the server job J1 repeats the above processing until receiving the end of the trace.

When the server job J1 receives the full notice of the trace buffer B1 from the virtual machine control system AVM, the server job J1 clears the empty buffer to the virtual machine control system AV.
Notify M. Then, the data of the filled buffer is output to the external storage medium D. Since the trace buffer B1 is composed of two buffers as described above, the storage of the trace data and the output thereof are performed in parallel.

As described above, in the present embodiment, the server jobs J1, J2, J3 corresponding to the respective virtual machines VM1 to VM3 are provided on the server virtual machine SVM, and the trace data is assigned to the respective virtual machines VM1 to VM. Since the server job is stored in the trace buffer of the server job, it is possible to measure a plurality of virtual machines with one server virtual machine SVM, and it is possible to eliminate the limitation of the number of measurement steps due to the buffer capacity as in the conventional example. it can. B. Spatial information (system-specific information) collection mechanism and analysis function When a hardware instruction that causes a space switch in the measured virtual machine is executed, space identification information (space ID, process ID, etc.) in the control register is collected. If
Hardware instructions executed thereafter can be analyzed for each space.

In the present embodiment, the instruction tracing mechanism of the virtual machine control system makes it possible to collect space identification information in the control register at the time of executing a hardware instruction which may cause space switching as described above. This is to identify the space to be created. FIG. 3 is a diagram for explaining the space information collecting mechanism described above. As shown in FIG. 3, after the trace in the virtual machine to be measured VM starts,
When the instructions (2) and (3) that cause a space switch are executed, the virtual machine control system AVM outputs the space identification information in the control register to the trace buffer of the server job.

In the server job, the space identification information described above is received, and the travel command of the space collected by the command (1), the travel command of the space collected by the command (2), and the space command collected by the command (3). Identify the travel command. As described above, in the present embodiment, when a hardware instruction that causes space switching is executed in the virtual machine control system, a mechanism for collecting space identification information in the control register is provided. The number of steps can be grasped. C. Collection mechanism and analysis function of dynamically loaded module information This embodiment is called when a space is created on the virtual machine to be measured, and “registers the address of the following loading completion exit routine in the current space. Routine (referred to as a space creation exit routine) "and a routine which is called when the loading of a module is completed and issues an instruction to store module information, load module information in a register, and store the contents of the register in a trace buffer. (Referred to as a loading completion exit routine) "to collect information on dynamically loaded modules, whereby the number of running steps for each module in each space can be ascertained. . The function achieved by the space creation exit routine and the loading completion exit routine is called an exit function.

In this embodiment, module information is collected as follows. At the start of measurement, dynamic module information (region / module information) of the entire space is collected. At this time, a loading completion exit is registered for all spaces. That is, the address for calling the above-mentioned loading completion exit routine is registered.

When a new space is created during measurement, the above-mentioned space creation exit routine is called, and the above-mentioned loading completion exit is registered in the created space by the space creation exit routine. Then, if a module is loaded during the measurement, the loading completion exit routine is called, and the loading completion exit routine stores the information of the loading module. The stored information of the loading module is output to the trace buffer of the server job using the interface with the virtual machine control system.

FIG. 4 is a diagram for explaining a mechanism for collecting dynamic module information according to the present embodiment. The mechanism for collecting dynamic module information according to the present embodiment will be described with reference to FIG. Note that the numbers in parentheses (1), (2),... In FIG. (1) Start tracing in the measured virtual machine VM. (2) When the space A is newly created, the space creation exit routine is called from the space A. At this time, space switching information is output to the trace buffer of the server job. (3) The space creation exit routine registers a loading completion exit in the created current space A. That is, the address for calling the loading completion exit routine is registered as described above. After that, the virtual machine VM to be measured creates a space specific area, and outputs a running step of the OS for space creation to the trace buffer of the server job until the loading is completed. (4) When the loading of the load module is completed in the space A, the loading completion exit routine is called in the virtual machine to be measured. (5) The loading completion exit routine stores the loaded module information. (6) Next, the loading completion exit routine places the module information in the register and issues the BCRO, 1 instruction. BCR
The O, 1 instruction is a hardware instruction that outputs the contents of the register. By issuing this instruction, the loading module information is output to the server job trace buffer using the interface of the virtual machine control system AVM. You.

Thereafter, when the loading module is executed in the virtual machine VM to be measured, the running step of the loading module is output to the trace buffer of the server job. Each of the above-mentioned exit routines runs only about 3K steps per call, and operates only at the time of module loading and space creation, so that the overhead during measurement is very small.

FIG. 5 shows a static module at the start of tracing,
FIG. 3 is a diagram for explaining a dynamic module information collection mechanism. The following control is performed on a space existing before the start of measurement. In addition, similarly to FIG.
(1), (2), ... correspond to the following numbers in parentheses. (1) On the virtual machine to be measured, start a measurement setup job and schedule the SRBR. That is, SRB
According to the schedule, the program is operated in the spaces A and B of the schedule destination to perform the activity process. (2) In the spaces A and B, the activity sets the module information of its own space in the module information management table. The module information management table is a table for holding unique information on the virtual machine to be measured, and based on the information registered in this table, after the measurement is completed, detailed system information on the virtual machine to be measured is collected. (3) Next, in the above activity, a loading completion exit is registered. (4) The measurement setup job outputs the contents of the module management information table to the external medium after the activities of all spaces are completed. The data can also be passed to a server job on the server virtual machine SVM by using the inter-virtual machine communication function of the virtual machine control system AVM and other communication functions. (5) After the start of the instruction trace, module loading occurs. When the loading is completed, the loading completion exit routine is called as described above. (6) The loading completion exit routine stores the module information as described above. (7) Next, the loading completion exit routine places the module information in the register as described above, issues a BCRO, 1 instruction, and traces the stored module information using the interface of the virtual machine control system to trace the server job. Output to buffer.

As described above, in this embodiment, the exit function called when the space is created and the exit function called when the loading of the module is completed are provided. Information on dynamically loaded modules can be accurately collected, and the number of running steps for each module in each space can be ascertained. D. Collection mechanism of other system specific information C. Has an interface between the virtual machine control system and the virtual machine to be measured using specific hardware instructions to collect information on dynamically loaded modules, but uses a similar mechanism. This allows other system-specific information to be output as instruction trace data with minimum overhead.

As described above, this embodiment has an interface with the virtual machine control system with one hardware instruction and collects register information at the time of execution of the instruction. Collection of unique information is performed as follows. That is, when information unique to the system is stored in the register, the measured virtual machine issues a BCRO, 1 instruction, which is a hardware instruction, and outputs necessary information to the trace buffer of the server job.

The person who executes the hardware of the interface can understand the meaning of the information stored in each register, and can determine the hardware instruction of the interface executed by another person based on the information stored in the register. At the time of information analysis, necessary information is extracted from the register of the hardware instruction of the interface executed by the user, and each user can perform its own analysis.

FIG. 6 is a diagram showing a mechanism for collecting other unique information according to this embodiment. In this embodiment, other unique information is collected as follows. In the same manner as above, the numbers in parentheses (1), (2),... In the figure correspond to the numbers in parentheses below. (1) Instruction tracing is started in the measured virtual machine VM. Thereby, the instruction trace data is output to the trace buffer of the server job J. (2) A hardware instruction to be executed on the virtual machine VM to be measured is output to a trace buffer prepared by the server job SVM. (3) Execute the hardware instruction of the interface. (4) If the hardware instruction executed on the virtual machine VM to be measured is an interface instruction, register information is collected after the execution of the instruction (● in FIG. 6). (5) The hardware instruction to be executed on the virtual machine VM to be measured is output to the buffer prepared by the server job J. (6) ● The register information at the time is output to the buffer prepared by the server job J. E. FIG. Ascertaining the number of dynamic steps in a plurality of virtual spaces As in UNIX described above, even in an OS in which a hardware instruction for switching virtual spaces cannot be specified, an OS
When switching the virtual space or the like by the control program (kernel) of D. By collecting the space identification information using the interface described in (1), the number of dynamic steps for each virtual space can be grasped.

As described above, in an OS that cannot specify a hardware instruction for switching the virtual space,
B. O cannot be used because the mechanism shown in FIG.
The same operation is performed on the program side of S. Will be used. As described above, when the space is switched, the present embodiment collects information that the space has been switched and information (space identification information) on which memory map to use, and performs switching of the virtual space. The present embodiment enables the number of dynamic steps in a plurality of virtual spaces to be ascertained in an OS in which a hardware instruction to be executed cannot be specified. This embodiment will be described with reference to FIG. (1) In the virtual machine to be measured VM, instruction tracing is started and instructions are executed in the kernel space. The trace data is output to a server job trace buffer of the server virtual machine SVM via the virtual machine control system AVM. Then, it is determined from the program counter information that the vehicle is traveling in the kernel space, and the analysis function counts the number of steps for each module in the kernel space. (2) The virtual space is switched in the measured virtual machine VM, and the interface command is executed. The virtual machine control system AVM collects identification information on the space A according to the interface command, and the space identification information is output to the trace buffer of the server virtual machine SVM. Then, the trace data in the space A of the virtual machine VM to be measured is output to the trace buffer of the server virtual machine SVM, and the number of steps for each module in the space A is counted by the analysis function.

When the processing of the space A in the virtual machine VM to be measured is completed and the process returns to the kernel space, it is determined from the program counter information (PSW) that the control program of the OS is running. The number of steps for each module in the space is counted. (3) In the virtual machine to be measured VM, the virtual space is switched again, and the virtual machine control system AVM collects the identification information on the space B and counts the number of steps for each module in the space B as described above. . Further, when the processing of the space B in the virtual machine VM to be measured is completed and the process returns to the kernel space, it is determined from the program counter information (PSW) that the OS control program is running, and a step for each module in the kernel space is performed. The number is counted.

As described above, in the present embodiment, the above-described D.E. Since the space identification information is collected by using the above interface, the number of dynamic steps in a plurality of virtual spaces can be grasped in an OS in which a hardware command for switching virtual spaces cannot be specified. F. Optimization of space identification information Information on which file is used to obtain a memory map must be collected as space identification information. For example, in an OS such as UNIX, file identification is generally called a path name. The name is determined by the name of the file, and the length of the path name is not fixed.

On the other hand, the path name in the UNIX system is a logically assigned identification name. Inside the system, a file is identified by three kinds of identifiers called a major number, a minor number, and an inode number indicating the physical location of the file. Is managing. In the present embodiment, a major number, a minor number, i
The memory map is created by storing three types of identifiers of the node numbers, thereby preventing an increase in the amount of stored data.

That is, as shown in FIG. 8, the identifier of program A in UNIX is "/ usr / bin / sh", and the identifier of program B is "/ usr / ccs / bin / cc".
In the case of "", 12 bytes and 16 bytes are required, respectively, because a null indicating the end is entered at the end. However, the use of three types of identifiers of a major number, a minor number, and an inode number fixes the number to 12 bytes.

As described above, in this embodiment, since three types of identifiers, that is, the major number, the minor number, and the inode number, are used, the amount of data for storing information as to which memory map is used. Can be prevented from increasing. UNIX is a registered trademark in the United States and other countries licensed by X / Open Company Limited.

[0055]

As described above, the following effects can be obtained in the present invention. According to the first aspect of the present invention, a virtual machine control system
Instruction tray for each virtual machine to be raced
Source via the virtual machine control system,
This collection result is used according to the traced virtual machine.
Since the server virtual machine for outputting to the trace buffer is provided, it is possible to measure a plurality of virtual machines with one server virtual machine, and it is possible to eliminate the limitation of the number of measurement steps due to the capacity of the buffer as in the conventional example. .
According to a second aspect of the present invention, in the first aspect of the present invention, the first computer is provided for each of the virtual machines to be traced.
A second trace buffer and a second trace buffer.
The output of the output means is connected to the first buffer or the first buffer.
Is applied to one of the second buffers and
The first battery not outputting to the output means.
Or the tray in which the second buffer is also stored.
Since the second output means for outputting the source data to the external medium is provided, the data in the trace buffer can be output to the external medium while the trace data is accumulated, and the trace data cannot be output because the buffer becomes full. Can be prevented. In the invention of claim 3,
In the invention of claim 1 or claim 2, the server temporary
Since the virtual computer is provided with a mechanism for collecting space identification information in the control register and outputting it as trace data when executing a hardware instruction that causes space switching , the number of running steps can be ascertained for each space. Claim 4
In the invention, the virtual machine to be traced according to any one of claims 1, 2 and 3,
A first exit function that is called when a space is created and registers a loading completion exit in the created space, and stores module information when module loading is completed, and stores the module information in a register. ,
Since a second exit function for outputting as trace data using the interface of the virtual machine control system is provided, it is possible to accurately collect information on dynamically loaded modules with minimum overhead. It is possible to grasp the number of traveling steps for each module in the space. In the invention of claim 5, claim 1,
In the second, third or fourth aspect of the present invention, tracing is performed.
A measurement setup job is provided in the virtual machine to be processed, and the module information of the space where the measurement setup job exists before the trace starts is set in the management table, and the loading completion exit is registered. Static module and dynamic module information at the start of tracing can be collected. Claim 6
In the invention, in the invention of claim 1, 2, 3, 4 or claim 5, a mechanism that outputs register information at the time of hardware instruction execution as trace data, Torre
On the virtual machine to be executed, when necessary information specific to the system is provided in the register,
Since the system is configured to collect necessary system-specific information by executing the above hardware instructions, system-specific information can be collected with a minimum overhead. In the invention of claim 7, claims 1, 2, and
In the third, fourth, fifth, or sixth aspect of the present invention, when measuring a virtual machine having an OS that cannot specify a hardware instruction for switching a virtual space, a trace is collected to collect data and system-specific information. Subject to do
When said hardware instruction is executed in the virtual machine that, the mechanism for outputting the identification information of the switched space as trace data is provided, the space identification information output by executing said hardware instruction when space switching Since the configuration is such that the number of dynamic steps for each virtual space is ascertained, the number of dynamic steps in a plurality of virtual spaces can be ascertained in an OS in which a hardware command for switching virtual spaces cannot be specified. In the invention of claim 8, claims 1, 2, 2
3. The method according to claim 3, wherein a variable-length identifier is used as information for identifying the file.
When measuring the virtual machine having S, instead of using the variable-length identifier, a fixed-length identifier indicating the physical location of the file is used and the fixed-length identifier is stored. It is possible to prevent an increase in the amount of data when storing information on whether to use the memory map.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating the principle of the present invention.

FIG. 2 is a diagram showing an instruction trace and collection mechanism of a plurality of virtual machines to be measured.

FIG. 3 is a diagram showing a spatial information collecting mechanism.

FIG. 4 is a diagram showing a dynamic module information collection mechanism using a space creation exit.

FIG. 5 is a diagram showing a static and dynamic module information collection mechanism at the start of a trace.

FIG. 6 is a diagram showing another system-specific information collection mechanism.

FIG. 7 is a diagram showing a dynamic step measurement mechanism for each space.

FIG. 8 is a diagram showing a method of optimizing space identification information.

FIG. 9 is a diagram for explaining a conventional instruction tracing mechanism.

FIG. 10 is a diagram showing a conventional tracing mechanism.

[Explanation of symbols]

AVM virtual machine control system SVM server virtual machine J1 to J3 server job VM, VM1 to VM3 virtual machine to be measured B1 to B3 trace buffer D External storage medium. AS analysis system

Continuation of front page (56) References JP-A-62-52644 (JP, A) JP-A-59-22149 (JP, A) JP-A-63-201840 (JP, A) JP-A-63-291139 (JP, A) , A) JP-A-63-20549 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G06F 11/28 310 G06F 9/46 350

Claims (8)

(57) [Claims] In 1. A virtual computer system for a plurality of instructions trace provisional <br/> virtual machine under virtual machine control system
There is a trace under the virtual machine control system.
The instruction trace for each virtual machine to be executed
Collected via the virtual machine control system,
The collected results are stored in a tray corresponding to the traced virtual machine.
A virtual machine system comprising a server virtual machine that outputs data to a buffer . 2. The method according to claim 1, wherein the trace buffer is a trace target.
A first trace buffer for each of the virtual machines
And a second trace buffer, wherein the output
The output of the means is either the first buffer or the second buffer.
The first buffer which is performed to any of the buffers and does not output to the output means.
Buffer or the buffer in which the second buffer is also stored.
Second output means for outputting race data to an external medium.
Virtual computer system according to claim 1, characterized in that it has. Wherein said server virtual machine, when the hardware instruction execution space switching occurs, collect the space identification information in the control register, claim and having a mechanism for outputting as trace data 1 or The virtual computer system according to claim 2. 4. The virtual meter to be traced.
A first exit function that is called when a space is created and registers a loading completion exit in the created space; and stores module information when module loading is completed, and stores the module information in a register. And a second exit function for outputting as trace data using the interface of the virtual machine control system to collect dynamically loaded module information. Item 3. The virtual computer system according to item 3. 5. A measurement setup job is provided in the virtual machine to be traced, module information of a space where the measurement setup job exists before the start of tracing is set in a management table, and a loading completion exit is set. The virtual computer system according to claim 4, wherein the registration is performed. 6. A mechanism for outputting register information at the time of execution of a hardware instruction as trace data is provided. When necessary information specific to a system is contained in a register on the virtual machine to be traced , By executing the above hardware instruction,
6. A system according to claim 1, wherein necessary system-specific information is collected.
Virtual computer system. 7. A virtual machine having an OS that cannot specify a hardware instruction for switching a virtual space.
A virtual machine system having a mechanism for outputting identification information of a switched space as trace data when a hardware instruction is executed in the virtual machine to be traced, wherein the hardware is provided at the time of space switching. 7. The system according to claim 1, further comprising a mechanism for grasping the number of dynamic steps for each virtual space based on space identification information output by executing the instruction. Virtual computer system. 8. The virtual computer system having a virtual machine length of DN as information for identifying the file to mount an OS to be used, instead of the variable length DN, physical files The information for forming a memory map is secured by using a fixed-length identifier indicating a location and storing the fixed-length identifier. Item 7. The virtual computer system according to Item 7.