JP5013099B2 - Disk bandwidth calculation system, method, and program - Google Patents

Description

  The present invention relates to a technique for calculating a disk bandwidth when access occurs to a plurality of files on a disk.

  For the purpose of improving the usage rate of high-performance computers, the introduction of virtualization technology is being studied. According to the virtualization technology, at least one virtual machine (VM) is built on a physical machine. More specifically, each virtual machine is assigned a physical machine resource such as a CPU, memory, and disk, and each virtual machine can freely use the assigned resource, as if one physical machine So that the software can be run. By sharing physical machine resources, a plurality of virtual machines can be operated on one physical machine.

  In a system in which such a virtualization technology is introduced, how to allocate resources to each virtual machine is important for improving resource utilization efficiency.

  Non-Patent Document 1 describes a method for allocating resources such as a CPU, a disk, and a network to a virtual machine. Regarding the CPU, the response time is calculated from the queue based on the load amount, and the CPU resource amount satisfying the requested response time is allocated. Regarding the disks, disk resources are allocated so that the sum of the requested disk sizes of each virtual machine does not exceed the disk size (disk capacity) of the physical machine. With respect to the network, network resources are allocated so that the sum of the required bandwidths of each virtual machine does not exceed the bandwidth of the physical machine.

Y. Ding and E. Bolker, “How many guests can you serve?-On the number of partitions”, the Proc. Of CMG conference, 2006.

The inventor of the present application paid attention to the following points.
As an example, consider a case where accesses are concentrated from a plurality of virtual machines on one disk in actual operation. In this case, there is a possibility that the “disk bandwidth”, which is the data transfer rate per unit time between the disk and the CPU, decreases. However, in the technique described in Non-Patent Document 1 described above, when determining the allocation of the disk resources to the virtual machine, only the disk capacity is considered, and the disk bandwidth is not considered. Therefore, after starting a plurality of virtual machines, if accesses from the virtual machines to the disks are concentrated, there is a possibility that a desired disk bandwidth cannot be obtained.

  Thus, it is preferable to consider the disk bandwidth when determining the resource allocation to the virtual machines. Therefore, a technique that can estimate the disk bandwidth when a disk is accessed from a plurality of virtual machines is desired. More generally, a technique that can estimate the disk bandwidth when a plurality of files on the disk are accessed is desired.

  One object of the present invention is to provide a technique capable of estimating the disk bandwidth when access occurs to a plurality of files on a disk.

  In a first aspect of the present invention, a disk bandwidth calculation system is provided that calculates a disk bandwidth when a plurality of files on a disk are accessed. The disk bandwidth calculation system includes a storage device, a representative seek distance calculation module, and a bandwidth calculation module. The storage device stores at least disk profile information indicating a statistical relationship between the seek distance and the disk bandwidth. The representative seek distance calculation module calculates a representative seek distance when an access occurs to a plurality of files. The bandwidth calculation module refers to the disk profile information and calculates the disk bandwidth corresponding to the representative seek distance as the disk bandwidth when access to a plurality of files occurs.

  In a second aspect of the present invention, there is provided a disk bandwidth calculation system for calculating a disk bandwidth when access is made to a plurality of files on a disk. The disk bandwidth calculation system includes a storage device and a disk bandwidth calculation module. The storage device stores disk profile information indicating a statistical relationship between seek distance and disk bandwidth, and seek distance information indicating a seek distance between different files and within the same file among a plurality of files. Is done. The disk bandwidth calculation module refers to the disk profile information and seek distance information, and calculates the disk bandwidth when an access occurs to a plurality of files.

  In a third aspect of the present invention, there is provided a disk bandwidth calculation method for calculating a disk bandwidth when access is made to a plurality of files on a disk. The disk bandwidth calculation method includes (A) a step of reading disk profile information indicating a statistical relationship between a seek distance and a disk bandwidth from a storage device, and (B) when a plurality of files are accessed. A step of calculating a representative seek distance; and (C) a disk bandwidth corresponding to the representative seek distance with reference to the disk profile information, and a disk bandwidth when a plurality of files are accessed. Calculating as a width.

  In a fourth aspect of the present invention, there is provided a disk bandwidth calculation method for calculating a disk bandwidth when access is made to a plurality of files on a disk. The disk bandwidth calculation method includes (A) disk profile information indicating a statistical relationship between the seek distance and the disk bandwidth, and a seek that indicates a single seek distance between different files and within the same file. Reading distance information from the storage device; and (B) calculating disk bandwidth when access to a plurality of files occurs with reference to disk profile information and seek distance information. .

  In a fifth aspect of the present invention, a disk bandwidth calculation program for calculating a disk bandwidth when access to a plurality of files on a disk occurs is provided. The disk bandwidth calculation program causes the computer to execute the following processing. That is, the processing includes (a) a step of reading disk profile information indicating a statistical relationship between the seek distance and the disk bandwidth from the storage device, and (b) a representative example when access to a plurality of files occurs. And (c) referring to the disk profile information, the disk bandwidth corresponding to the representative seek distance is determined as the disk bandwidth when a plurality of files are accessed. Calculating.

  In a sixth aspect of the present invention, a disk bandwidth calculation program for calculating a disk bandwidth when access to a plurality of files on a disk occurs is provided. The disk bandwidth calculation program causes the computer to execute the following processing. That is, the processing includes (a) disk profile information indicating a statistical relationship between seek distance and disk bandwidth, and seek distance indicating one seek distance between different files and within the same file among the plurality of files. Reading information from the storage device; (b) referring to the disk profile information and the seek distance information; calculating a disk bandwidth when an access occurs to the plurality of files; including.

  According to the present invention, it is possible to estimate the disk bandwidth when access is made to a plurality of files on a disk.

  When reading or writing certain data on the disk, the disk head moves to a position corresponding to that data. For example, the disk head reads data for one read size and then repeats movement within the same file or between different files. Such movement of the disk head is called “seek”.

  First, a case where access occurs to a plurality of files on a disk will be described with reference to FIG. As an example, consider a case where n virtual machines VM1 to VMn (n is an integer of 2 or more) operate on a physical machine. The physical machine has a disk 100, and different areas of the disk 100 are assigned to each of the n virtual machines VM1 to VMn. Here, consider a case where accesses from the n virtual machines VM <b> 1 to VMn are concentrated on the disk 100. For example, read accesses occur almost simultaneously from the virtual machines VM1 to VMn to the files F1 to Fn on the disk 100, respectively. In this case, the disk head reads the data of each file while reciprocating between the n files F1 to Fn. Therefore, a decrease in disk bandwidth occurs.

  As illustrated in FIG. 1, a state in which access to a plurality of files F1 to Fn on the disk 100 occurs can be considered. Such a state may be referred to as “multiple file access” hereinafter. The embodiment of the present invention provides a technique for estimating the disk bandwidth when multiple file accesses occur. Hereinafter, various embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1. 1. First embodiment 1-1. Configuration FIG. 2 is a block diagram showing the configuration of the disk bandwidth calculation system 1 according to the first embodiment. The disk bandwidth calculation system 1 is a computer system for estimating the disk bandwidth when multiple file accesses occur. The disk bandwidth calculation system 1 includes a storage device 2, a processing device 3, an input device 4, and an output device 5. Examples of the storage device 2 include an HDD (Hard Disk Drive) and a RAM (Random Access Memory). The processing device 3 includes a CPU (Central Processing Unit). Examples of the input device 4 include a keyboard, a mouse, a media drive, and a network interface. Examples of the output device 5 include a display, a media drive, and a network interface.

  The storage device 2 stores file position information LOC, access characteristic information ACS, disk profile information PRF, and seek distance information SEK. The details of each information are as follows.

(File location information LOC)
The file position information LOC is information indicating the position on the disk 100 of the files F1 to Fn to be accessed. FIG. 3 shows two files Fi and Fj among the files F1 to Fn on the disk 100 (i = 1 to n, j = 1 to n, i ≠ j). As shown in FIG. 3, for each file, a start position and an end position on the disk 100 can be defined. The file position information LOC is information indicating the start position and end position of each file. Alternatively, since the file size is defined by the size from the start position to the end position, the file position information LOC may indicate either the start position or the end position and the file size.

  FIG. 4A shows an example of the file position information LOC. In the example shown in FIG. 4A, the file position information LOC indicates the correspondence between the file name, the start position, and the end position for each of the files F1 to Fn. FIG. 4B shows another example of the file position information LOC. In the example shown in FIG. 4B, the file position information LOC indicates the correspondence between the file name, the start position, and the file size for each of the files F1 to Fn. The end position of each file can be obtained from the start position and the file size. In the example shown in FIGS. 4A and 4B, the number of bytes from the top of the disk 100 is used as a unit representing the file position, but is not limited thereto. The file position may be represented by the number of cylinders from the top of the disk 100.

(Access characteristic information ACS)
The access characteristic information ACS is information indicating the characteristics of file access in the disk 100. In other words, the access characteristic information ACS indicates the change pattern of the access target file on the disk 100. The change pattern of such an access target file is determined by a file access scheduling policy. In general, the file access scheduling policy differs depending on platforms such as Linux and Xen virtual machines. Therefore, the change pattern of the access target file also differs depending on the platform type.

The access characteristic information ACS gives, for example, “file switching rate α”. The file switching rate α will be described with reference to FIG. Consider a case where read access occurs for the two files Fi and Fj shown in FIG. In this case, the disk head repeats the operation of reading and moving the data of the file Fi or Fj by one read size. At this time, the probability that the data of the file Fj is read after the file Fi is the file switching rate α _ij . Similarly, the probability that the data of the file Fi is read after the file Fj is the file switching rate α _ji . The disk head may read data of the same file again after a certain file. The probability that the data of the same file Fi is read next to the file Fi is the file switching rate α _ii , and the probability that the data of the same file Fj is read next to the file Fj is the file switching rate α _jj . More generally, the probability that the second file is accessed next to the first file among the plurality of files on the disk 100 is the “file switching rate between the first file and the second file. α ”. The first file and the second file may be the same. The file switching rate α can be obtained from the file access history during actual system operation.

FIG. 5A shows an example of the access characteristic information ACS. In the example shown in Figure 5A, the access characteristic information ACS shows the probability matrix about the file switch rate alpha _ij, file switching rate _{α ij (i = 1~n, j} = 1~n) between each file matrix Is given in the form. FIG. 5B shows another example of the access characteristic information ACS. In the example shown in FIG. 5B, only the probability α _ii (i = 1 to n) that the same file is continuously accessed is shown. In this case, it is assumed that the probability α _ij (i = 1 to n, j = 1 to n, i ≠ j) that the next different file is accessed after a certain file is all equal. That is, the file switching rates α _ij are all assumed to be the same value (1−α _ii ) / (n−1).

(Disc profile information PRF)
The disk profile information PRF is information indicating a statistical relationship between the seek distance of the disk head and the disk bandwidth.

  In order to explain in detail an example of the correspondence between the seek distance and the disk bandwidth, consider a case where N read accesses are continuously performed on a certain test area in the disk. At this time, the disk head performs N seek operations in the test area. When N is a large number and N read accesses occur randomly in the test area, the average value x of the N seek distances is half the size LT of the test area (x = LT / 2). On the other hand, assume that the amount of read data per time is SR, and the time required for N times of read access is T. At this time, the disk bandwidth y per unit time is given by y = (N × SR) / T. In this way, a certain average seek distance x (= LT / 2) and the corresponding disk bandwidth y are obtained. Various correspondences between the average seek distance x and the disk bandwidth y, that is, the average seek distance x and the disk bandwidth y are set by variously setting the test area size LT within the range of the total size of the disk. And get a statistical relationship.

  FIG. 6A is a graph showing the correspondence relationship y = f (x). In FIG. 6A, the horizontal axis represents the average seek distance x, and the vertical axis represents the disk bandwidth y. As shown in FIG. 6A, the shorter the average seek distance x, the larger the disk bandwidth y, while the longer the average seek distance x, the smaller the disk bandwidth y. This relationship has also been reported in C. Ruemmler and J. Wilkes, “An Introduction to Disk Drive Modeling”, Computer, Vol. 27, No. 3, pp. 17-29, 1994.

  The disk profile information PRF indicates, for example, a statistical correspondence between the average seek distance and the disk bandwidth described above. For example, the disc profile information PRF gives a function y = f (x) as shown in FIG. 6A. Alternatively, as shown in FIG. 6B, the disk profile information PRF may give the correspondence between the average seek distance and the disk bandwidth in a table format. According to FIG. 6B, for example, the disk bandwidth is 0.5 Mbyte / sec when the average seek distance is 0.5 Gbyte, and the disk bandwidth is 0.23 Mbyte / sec when the average seek distance is 64 Gbyte. I understand. The disk bandwidth corresponding to the average seek distance not directly shown in the table entries can be calculated by interpolating between the entries. The seek distance can be expressed not only by the number of bytes but also by the number of cylinders.

  Further, the disk profile information PRF is not limited to the above-described example, and it is only necessary to indicate the correspondence relationship between the seek distance and the disk bandwidth obtained statistically or empirically. For example, the disk profile information PRF may indicate a correspondence relationship between the median value of the seek distance obtained statistically and the disk bandwidth.

(Seek distance information SEK)
The seek distance information SEK is information that gives an index of one seek distance. Details of the seek distance information SEK will be described later.

  Referring to FIG. 2 again, the storage device 2 further stores a program PROG. The program PROG is software executed by the processing device 3. Typically, the program PROG is recorded on a computer-readable recording medium and is read by the processing device 3. Alternatively, the program PROG may be provided via a network.

  The processing device 3 provides various functions by executing the program PROG. Specifically, the processing device 3 includes a data input module 10, a seek distance calculation module 30, a disk bandwidth calculation module 40, and an output module 70. These modules are realized by the processing device 3 executing the program PROG. That is, these modules are realized by the cooperation of the processing device 3 and the program PROG. Hereinafter, processing by each module will be described in detail.

1-2. Processing Flow FIG. 7 is a block diagram showing a processing flow in the first embodiment. FIG. 8 is a flowchart showing the processing in the first embodiment.

(Step S10)
Various information necessary for the processing is input through the input device 4. The data input module 10 acquires various pieces of input information and stores them in the storage device 2. In the present embodiment, the data input module 10 acquires the above-described file position information LOC, access characteristic information ACS, and disk profile information PRF and stores them in the storage device 2.

(Step S30)
The seek distance calculation module 30 reads the file position information LOC from the storage device 2 and calculates “file seek distance L _ij ” and “file seek distance L _ii ” based on the file position information LOC. The inter-file seek distance L _ij and the intra-file seek distance L _ii will be described with reference to FIG.

The seek distance between files L _ij is one seek distance between two different files Fi and Fj among the files F1 to Fn to be accessed. For example, the seek distance between files L _ij is an average seek distance between the file Fi and the file Fj. In this case, as shown in FIG. 9, the inter-file seek distance L _ij is given by the distance between the center position of the file Fi and the center position of the file Fj. Alternatively, the inter-file seek distance L _ij may be a distance (longest distance) between the start position of the file Fi and the end position of the file Fj. Alternatively, the inter-file seek distance L _ij may be a distance (shortest distance) between the end position of the file Fi and the start position of the file Fj.

On the other hand, the in-file seek distance L _ii is a one-time seek distance within the same file Fi. For example, the file seek distance L _ii is an average seek distance in the file Fi. Assuming that the access in the file is random, the seek distance L _ii in the file is a value that is half the file size. Alternatively, the in-file seek distance L _ii may be given by the distance from the start position to the end position of the file Fi, that is, the file size of the file Fi.

Thus, the seek distance between files L _ij and the seek distance within file L _ii indicate an index of one seek distance, and can be calculated from the above-described file position information LOC (start position, end position, file size). It is. Information indicating the calculated inter-file seek distance L _ij and intra-file seek distance L _ii is “seek distance information SEK”. That is, the seek distance calculation module 30 creates seek distance information SEK based on the file position information LOC.

(Step S40)
The disk bandwidth calculation module 40 reads seek distance information SEK, access characteristic information ACS, and disk profile information PRF from the storage device 2. As described above, the access characteristic information ACS indicates the characteristics of file access such as the file switching rate α. Therefore, it is possible to consider the situation when multiple file accesses occur based on the access characteristic information ACS. The disk profile information PRF indicates a statistical correspondence between the average seek distance and the disk bandwidth. Accordingly, the disk bandwidth calculation module 40 can calculate the disk bandwidth when multiple file accesses occur by referring to the seek distance information SEK, the access characteristic information ACS, and the disk profile information PRF.

  For example, as shown in FIG. 7, the disk bandwidth calculation module 40 includes a representative seek distance calculation module 41 and a bandwidth calculation module 42. The representative seek distance calculation module 41 and the bandwidth calculation module 42 perform the following processing, respectively.

(Step S41)
The representative seek distance calculation module 41 refers to the seek distances L _ij and L _ii indicated by the seek distance information SEK and represents a representative seek distance (hereinafter referred to as “representative seek distance Lrep”) when multiple file accesses occur. Calculated).

In calculating the representative seek distance Lrep, the representative seek distance calculation module 41 can refer to the access characteristic information ACS. For example, it is assumed that the access characteristic information ACS indicates the file switching rate α. As shown in FIG. 9, the probability that the file Fj is accessed next to the file Fi is α _ij , and the seek distance in this case is L _ij . The probability that the same file Fi is accessed next to the file Fi is α _ii , and the seek distance in this case is L _ii . Therefore, the representative seek distance calculation module 41 can calculate the probable expected value (average value) of the seek distance when multiple file accesses occur by using the file switching rate α. The expected value is used as the representative seek distance Lrep. More specifically, the representative seek distance Lrep in this case is expressed by the following equation (1).

The representative seek distance Lrep given by the above equation (1) is a weighted average value of the seek distances L _ij and L _ii calculated in step S30. That is, the representative seek distance calculation module 41 calculates the weighted average value of one seek distance L _ij and L _ii as the representative seek distance Lrep using the file switching rate α _ij as a weighting coefficient.

(Step S42)
The bandwidth calculation module 42 refers to the disk profile information PRF and calculates a disk bandwidth BW corresponding to the representative seek distance Lrep calculated in step S41. When the disk profile information PRF gives a function y = f (x) as shown in FIG. 6A, the corresponding disk bandwidth BW can be calculated by substituting the representative seek distance Lrep into the function. . As shown in FIG. 6B, when the correspondence relationship between the average seek distance and the disk bandwidth is given in a table format, the disk bandwidth BW corresponding to the representative seek distance Lrep can be calculated by interpolation. Alternatively, the fitting function may be obtained based on the table, and the disk bandwidth BW may be calculated by using the fitting function. The disk bandwidth BW calculated in this way is an estimated value of the disk bandwidth when multiple file accesses occur.

(Step S70)
The output module 70 outputs the estimated disk bandwidth BW through the output device 5. For example, the output device 5 is a display, and the disk bandwidth BW is displayed on the display. The user can refer to the display to determine whether or not the disk bandwidth BW when a plurality of file accesses occurs satisfies the desired disk bandwidth. In other words, it can be determined whether or not a desired disk bandwidth is guaranteed even if multiple file accesses occur.

1-3. Example of Processing Next, an example of processing will be described using specific numerical values. Assume that the files F1 to Fn to be accessed are image files of the three virtual machines VM1 to VM3 operating on the physical machine (n = 3). For example, when the three virtual machines VM1 to VM3 are activated, access to each VM image file on the disk occurs.

FIG. 10 shows various information input in this example. The file position information LOC indicates the start position and end position on the disk 100 of each of the three VM image files. The access characteristic information ACS indicates only the file switching rate α _ii , that is, the probability α _ii that the same file is continuously accessed. In this case, it is assumed that the probabilities α _{ij for} accessing the next different file after a certain file are all equal. That is, the file switching rates α _ij (i ≠ j) are all assumed to be the same value (1−α _ii ) / (n−1). In this example, Xen is assumed as the platform, and the file switching rate α _ii is 0.0. This means that a different file is always accessed after a certain file. Since n = 3, the file switching rate α _ij (i ≠ j) is 0.5.

In step S30, the seek distance calculation module 30 calculates the inter-file seek distance L _ij and the intra-file seek distance L _ii based on the file position information LOC. In this example, the seek distance between files is calculated as L ₁₂ = L ₂₁ = 30 [GByte], L ₁₃ = L ₃₁ = 70 [GByte], and L ₂₃ = L ₃₂ = 40 [GByte]. Further, since the size of each image file is 10 [GByte], the seek distance within the file is calculated as L ₁₁ = L ₂₂ = L ₃₃ = 5 [GByte].

In step S41, the representative seek distance calculation module 41 calculates the representative seek distance Lrep according to the above equation (1). In this example, as shown below, the representative seek distance Lrep is calculated as 46.7 [GByte];
Lrep = (0 × 5 + 0.5 × 30 + 0.5 × 70
+ 0.5 × 30 + 0 × 5 + 0.5 × 40
+ 0.5 × 70 + 0.5 × 40 + 0 × 5) / 3
= 46.7

In step S42, the bandwidth calculation module 42 calculates a disk bandwidth BW corresponding to the representative seek distance Lrep. In this example, the disk profile information PRF is given in a table format, the disk bandwidth corresponding to the average seek distance = 32 [GByte] is 0.27 [MByte / sec], and the average seek distance = 64 [GByte]. ] Has a disk bandwidth of 0.23 [MByte / sec]. Accordingly, the disk bandwidth BW corresponding to the representative seek distance Lrep (= 46.7 [GByte]) is calculated to be about 0.252 [MByte / sec] by linear approximation. Further, since the file size and the file switching rate α _{ij of} each file are equal, the disk bandwidth of 0.252 [MByte / sec] is equally distributed, and the disk bandwidth BW for each file is 0.084 [MByte / sec. ] Is calculated.

  In step S70, the output module 70 outputs the calculated disk bandwidth BW. In this example, 0.084 [MByte / sec] is output as the disk bandwidth for each file. Also, 0.252 [MByte / sec] is output as the sum of the disk bandwidths related to all files.

1-4. Effect As described above, according to the present embodiment, it is possible to estimate the disk bandwidth BW when a plurality of files on the disk 100 are accessed. The user can determine whether or not the estimated disk bandwidth BW satisfies the desired disk bandwidth. In other words, it can be determined whether or not a desired disk bandwidth is guaranteed even if multiple file accesses occur.

  Further, according to the present embodiment, when calculating the representative seek distance Lrep, the file switching rate α depending on the file access scheduling policy is used as the weighting coefficient. The scheduling policy differs depending on platforms such as Linux and Xen virtual machines. By using the file switching rate α, it is possible to perform precise analysis according to the type of platform.

1-5. Modification (1)
As described above, in step S40, the disk bandwidth calculation module 40 calculates the disk bandwidth BW when multiple file accesses occur. At this time, the disk bandwidth calculation module 40 can also calculate the disk bandwidth BW according to the following equation (2).

  In this case, it is not necessary to calculate the representative seek distance Lrep. The disk bandwidth calculation module 40 can calculate the disk bandwidth BW by referring to the seek distance information SEK, the access characteristic information ACS, and the disk profile information PRF.

1-6. Modification (2)
The content indicated by the access characteristic information ACS is not limited to the file switching rate α. For example, the access characteristic information ACS may indicate an average number d of times that the same file is continuously accessed. Hereinafter, the number d is referred to as “the number of consecutive accesses”. This continuous access count d is also determined by the file access scheduling policy.

Consider the case where all the files have the same characteristics and there is no bias in the access rate for each file. In this case, the above-described file switching rate α _ii does not depend on the value of i, and the following relational expression: α _ii = d / (d + 1) holds between the file switching rate α _ii and the number of consecutive accesses d. In step S41, the representative seek distance calculation module 41 can calculate the representative seek distance Lrep according to the following equation (3).

Similar to the case of Expression (1), the representative seek distance Lrep given by Expression (3) can be said to be a weighted average value of the seek distances L _ij and L _ii calculated in Step S30. That is, the representative seek distance calculation module 41 can calculate the weighted average value of the seek distances L _ij and L _ii as the representative seek distance Lrep by using the continuous access count d.

  Alternatively, similarly to the modification (1), the disk bandwidth BW can be calculated without obtaining the representative seek distance Lrep. Specifically, the disk bandwidth calculation module 40 can calculate the disk bandwidth BW according to the following equation (4).

1-7. Modification (3)
FIG. 11 is a diagram similar to FIG. 7, and conceptually shows another processing flow. In the modification shown in FIG. 11, the representative seek distance calculation module 41 refers to the seek distance information SEK and calculates the “maximum value” of the seek distances L _ii and L _ij as the representative seek distance Lrep (step S41). Specifically, the maximum value of the inter-file seek distance L _ij is the representative seek distance Lrep. The bandwidth calculation module 42 calculates a disk bandwidth BW corresponding to the maximum value of the inter-file seek distance L _ij (step S42). The disk bandwidth BW obtained in this way corresponds to the “worst value” of the disk bandwidth when multiple file accesses occur.

  It should be noted that in this modification, it is not necessary to use the access characteristic information ACS for calculating the disk bandwidth BW. As shown in FIG. 11, the disk bandwidth calculation module 40 refers to only seek distance information SEK and disk profile information PRF to calculate the disk bandwidth BW when multiple file accesses occur. Therefore, the processing is simplified. Further, according to this modification, it is possible to estimate the worst value of the disk bandwidth when multiple file accesses occur.

1-8. Modification (4)
The seek distance information SEK that gives the seek distances L _ij and L _ii may be created in advance. The creation method of the seek distance information SEK is the same as the creation method by the seek distance calculation module 30 described above.

  FIG. 12 conceptually shows the processing flow in this modification. The data input module 10 ′ acquires seek distance information SEK, access characteristic information ACS, and disk profile information PRF and stores them in the storage device 2. The disk bandwidth calculation module 40 reads seek distance information SEK and disk profile information PRF from the storage device 2. The access characteristic information ACS may be read as necessary. Then, the disk bandwidth calculation module 40 refers to the information and calculates the disk bandwidth BW when multiple file accesses occur. The calculation method may be any of the methods described in the above embodiments and modifications.

2. Second Embodiment According to the technique described in Non-Patent Document 1 described above, only the disk capacity of a physical machine is considered when determining the allocation of disk resources to a virtual machine. Specifically, the disk resources are allocated so that the total required disk size of each virtual machine does not exceed the disk capacity. In this case, the maximum number of virtual machines that can be operated on one disk is estimated by dividing the disk capacity by the disk size allocated to one virtual machine. However, this method does not take into account a decrease in disk bandwidth when multiple file accesses occur. Therefore, after starting the estimated maximum number of virtual machines and concentrating accesses from the virtual machines to the disks, there is a possibility that a desired disk bandwidth cannot be obtained.

  Therefore, in the second embodiment of the present invention, a technique for estimating the maximum number of files that can guarantee a desired disk bandwidth even when multiple file accesses occur is proposed. The desired disk bandwidth to be guaranteed for a disk is hereinafter referred to as “guaranteed bandwidth”. As will be described below, the maximum number of files satisfying the guaranteed bandwidth can be calculated by applying the first embodiment described above.

  In the following description, as an example, consider a case where a plurality of virtual machines VM1 to VMn access a single disk 100 as shown in FIG. The maximum number of virtual machines that satisfy the guaranteed bandwidth even in that state is estimated. Hereinafter, the number n of the virtual machines VM1 to VMn is simply referred to as “VM number n”. The disk size allocated to one virtual machine is simply referred to as “VM size”. For example, when a virtual machine exists as an image file, the size of the VM image file corresponds to the VM size. When a plurality of virtual machines VM1 to VMn are activated, access to each VM image file on the disk 100 occurs.

2-1. Configuration FIG. 13 is a block diagram showing a configuration of a disk bandwidth calculation system 1a according to the second embodiment. The same components as those in the first embodiment are denoted by the same reference numerals, and redundant descriptions are omitted as appropriate.

  Compared to the first embodiment, the storage device 2 further stores VM size information VMS, disk size information DKS, and guaranteed bandwidth information GBW. The VM size information VMS indicates a VM size (VM image file size) related to one virtual machine. The disk size information DKS indicates the overall size (disk capacity) of the disk 100. The guaranteed bandwidth information GBW indicates the guaranteed bandwidth.

  In addition, the processing device 3 includes a data input module 10a, a file position information creation module 20, a seek distance calculation module 30, a disk bandwidth calculation module 40, a VM number determination module 50, and an output module 70a. These modules are also realized by the processing device 3 executing the program PROG. Hereinafter, processing by each module will be described in detail.

2-2. Processing Flow FIG. 14 is a block diagram showing a processing flow in the second embodiment. FIG. 15 is a flowchart showing processing in the second embodiment. The description overlapping with the first embodiment is omitted as appropriate.

(Step S10a)
The data input module 10 a acquires VM size information VMS, disk size information DKS, access characteristic information ACS, disk profile information PRF, and guaranteed bandwidth information GBW, and stores them in the storage device 2. As will be described later, the file position information LOC is created by the file position information creation module 20.

(Step S51)
The VM number determination module 50 (file number determination module) is a module that variably sets the VM number n (file number n). In step S51, the VM number determination module 50 sets the VM number n to an initial value. The initial value is 1, for example. When n = 1, multiple file access does not occur, but the initial value of the VM number n may be set to 1 for convenience in the processing of this embodiment. Of course, the initial value may be set to 2 or more.

  Thereafter, based on the VM number n set by the VM number determination module 50, the file position information creation module 20, the seek distance calculation module 30, and the disk bandwidth calculation module 40 execute their respective processes. Then, according to the processing result, the VM number determination module 50 changes the VM number n. That is, a loop process is executed for each VM number n set by the VM number determination module 50.

(Step S20)
The file position information creation module 20 creates file position information LOC corresponding to the VM number n set by the VM number determination module 50. For this purpose, the file position information creation module 20 reads the VM size information VMS and the disk size information DKS from the storage device 2. Then, the file location information creation module 20 assumes the placement location on the disk of the image file of each virtual machine based on the set number of VMs n, the VM size, and the disk size, and creates the file location information LOC. At this time, since the access bandwidth improves as the seek distance between files becomes shorter (see FIG. 6A), it is preferable that the arrangement position is determined so that the distance between the files becomes as short as possible. For example, VM image files for the number of set VMs n are arranged in order from the top of the free area of the disk 100. The file position information LOC created in this way indicates the start position, end position, or size of each VM image file. The created file position information LOC is stored in the storage device 2.

(Step S30)
As in the first embodiment, the seek distance calculation module 30 creates seek distance information SEK based on the file position information LOC created in step S20.

(Step S40)
Similar to the first embodiment, the disk bandwidth calculation module 40 calculates the disk bandwidth BW when multiple file accesses occur. Any of the above-described methods may be used as the method for calculating the disk bandwidth BW.

(Steps S52 to S54)
The VM number determination module 50 reads the guaranteed bandwidth information GBW from the storage device 2. Then, the VM number determination module 50 compares the disk bandwidth BW calculated in step S40 with the guaranteed bandwidth indicated by the guaranteed bandwidth information GBW (step S52). Then, based on the comparison result, the VM number determination module 50 changes the set VM number n.

  Specifically, the VM number determination module 50 determines whether or not the calculated disk bandwidth BW satisfies the guaranteed bandwidth. When the disk bandwidth BW satisfies the guaranteed bandwidth, that is, when the disk bandwidth BW is larger than the guaranteed bandwidth (step S53; No), the VM number determination module 50 increases the set VM number n (step S54). ). For example, the VM number determination module 50 increases the set VM number n by one. Thereafter, the process returns to step S20 described above. That is, the file position information creation module 20 updates the file position information LOC in accordance with the new set VM number n, the seek distance calculation module 30 creates the seek distance information SEK again, and the disk bandwidth calculation module 40 The bandwidth BW is calculated.

  When the calculated disk bandwidth BW does not satisfy the guaranteed bandwidth, that is, when the disk bandwidth BW becomes smaller than the guaranteed bandwidth (step S53; Yes), the loop processing ends there. The VM number determination module 50 can determine the maximum VM number n (MAX) satisfying the guaranteed bandwidth based on the set VM number n at this time.

(Step S70a)
The output module 70 a outputs the determined maximum VM number n (MAX) through the output device 5. The user can design the system with reference to the maximum number of VMs n (MAX).

2-3. Example of Processing Next, an example of processing will be described using specific numerical values. A VM image file is considered as a file to be accessed. The access characteristic information ACS and the disk profile information PRF are the same as those shown in FIG. The VM size is 10 [GByte]. The disk size is 160 [GByte]. The guaranteed bandwidth is 0.270 [MByte / sec].

  In step S51, the initial value of the VM number n is set to 1 (n = 1).

If n = 1, file position information LOC indicating the arrangement position of one VM image file F1 is created in step S20. For example, the start position and end position of the VM image file F1 are 0 [GB] and 10 [GB], respectively. In step S30, the file in the seek distance _{L 11} is calculated to be half the VM size 5 [GByte]. In step S41, the representative seek distance Lrep is calculated according to the above equation (1). Here, when n = 1, since the same file is continuously accessed, the file switching rate α ₁₁ = 1 is specially considered regardless of the access characteristic information ACS. Therefore, the representative seek distance Lrep is calculated as 5 [GByte]. In step S42, the disk bandwidth BW corresponding to the representative seek distance Lrep is calculated as 0.347 [MByte / sec]. This disk bandwidth BW is larger than the guaranteed bandwidth (step S53; No). Therefore, the VM number n is increased and set to 2.

When n = 2, in step S20, file position information LOC indicating the arrangement positions of the two VM image files F1 and F2 is created. For example, the VM image file F2 is arranged subsequent to the VM image file F1, and the start position and end position thereof are 10 [GByte] and 20 [GByte], respectively. In step S30, the in-file seek distances L ₁₁ and L ₂₂ are calculated as 5 [GBytes], and the inter-file seek distances L ₁₂ and L ₂₁ are calculated as 10 [GBytes]. In step S41, a representative seek distance Lrep is calculated. In the access characteristic information ACS of this example, only the file switching rate α _ii = 0 is shown, and it is assumed that the file switching rates α _ij (i ≠ j) are all equal. Specifically, the file switching rate α _ij (i ≠ j) is given by (1−α _ii ) / (n−1) = 1 / (n−1). When n = 2, the file switching rates α ₁₂ and α ₂₁ are both 1.0. Therefore, the representative seek distance Lrep is calculated as 10 [GByte]. In step S42, the disk bandwidth BW corresponding to the representative seek distance Lrep is calculated as 0.33 [MByte / sec]. This disk bandwidth BW is larger than the guaranteed bandwidth (step S53; No). Therefore, the VM number n is increased and set to 3.

  When the same processing is repeated and the number of VMs n is set to 8, a representative seek distance Lrep = 30 [GBe] and a disk bandwidth BW = 0.275 [MByte / sec] are obtained. The disk bandwidth BW is larger than the guaranteed bandwidth as a dependency (step S53; No), and the VM number n is set to 9. When the number of VMs n is 9, a representative seek distance Lrep = 33.3 [GByte] and a disk bandwidth BW = 0.268 [MByte / sec] are obtained. At this time, the disk bandwidth BW is lower than the guaranteed bandwidth (step S53; Yes). That is, the disk bandwidth BW when multiple file accesses occur does not satisfy the guaranteed bandwidth. Therefore, the maximum number n (MAX) of virtual machines that satisfy the guaranteed bandwidth is estimated to be “8”.

  In step S70a, the maximum number n (MAX) = 8 of virtual machines satisfying the guaranteed bandwidth is output. In addition, the disk bandwidth BW = 0.275 [MByte / sec] when multiple file accesses occur may be output.

2-4. Effect As described above, the disk bandwidth calculation system 1a according to the present embodiment calculates the disk bandwidth BW while changing the set VM number n. This makes it possible to estimate the maximum number n (MAX) of virtual machines that satisfy the guaranteed bandwidth even when multiple file accesses occur. By considering the maximum number n (MAX) estimated in this way in the system design, it is possible to prevent a situation in which a desired disk bandwidth cannot be obtained after the virtual machine is started.

  In the above example, the case of a VM image file has been described, but the same applies to other cases. By calculating the disk bandwidth BW while changing the number n of files to be accessed, the maximum number n (MAX) that satisfies the guaranteed bandwidth can be estimated.

3. Third Embodiment The method for determining the maximum number n (MAX) that satisfies the guaranteed bandwidth is not limited to that described in the second embodiment.

3-1. Configuration FIG. 16 is a block diagram showing a configuration of a disk bandwidth calculation system 1b according to the third embodiment. The same components as those in the second embodiment are denoted by the same reference numerals, and redundant description is omitted as appropriate. In the present embodiment, the processing device 3 includes a data input module 10b, a file position information creation module 20, a seek distance calculation module 30, a disk bandwidth calculation module 40, a VM number determination module 50b, and an output module 70b. Yes. These modules are also realized by the processing device 3 executing the program PROG. Hereinafter, processing by each module will be described in detail.

3-2. Processing Flow FIG. 17 is a block diagram showing a processing flow in the third embodiment. FIG. 18 is a flowchart illustrating the processing according to the third embodiment. A duplicate description with the second embodiment is omitted as appropriate.

(Step S10b)
The data input module 10b acquires VM size information VMS, disk size information DKS, access characteristic information ACS, disk profile information PRF, and guaranteed bandwidth information GBW, and stores them in the storage device 2.

(Step S51b)
The VM number determination module 50b (file number determination module) is a module that variably sets the VM number n (number of files). In step S51b, the VM number determination module 50b sets the VM number n to an initial value. In this embodiment, the initial value of the VM number n is set to a “limit value” that can be arranged on the disk 100. For this purpose, the VM number determination module 50 b reads the VM size information VMS and the disk size information DKS from the storage device 2. Then, the VM number determination module 50b calculates the limit value of the number of VMs that can be arranged by dividing the disk size (disk capacity) by the VM size. Then, the VM number determination module 50b initializes the VM number n to the limit value. For example, when the VM size is 10 [GBytes] and the disk size is 160 [GBytes], the initial value of the VM number n is 16.

  Thereafter, based on the VM number n set by the VM number determination module 50b, the file position information creation module 20, the seek distance calculation module 30, and the disk bandwidth calculation module 40 execute their respective processes (steps S20 to S20). S40). Each processing is the same as that in the second embodiment.

(Steps S52b to S54b)
The VM number determination module 50 b reads the guaranteed bandwidth information GBW from the storage device 2. Then, the VM number determination module 50b compares the disk bandwidth BW calculated in step S40 with the guaranteed bandwidth indicated by the guaranteed bandwidth information GBW (step S52b). Then, based on the comparison result, the VM number determination module 50b changes the set VM number n.

  Specifically, the VM number determination module 50b determines whether or not the calculated disk bandwidth BW satisfies the guaranteed bandwidth. When the disk bandwidth BW does not satisfy the guaranteed bandwidth, that is, when the disk bandwidth BW is smaller than the guaranteed bandwidth (step S53b; No), the VM number determination module 50b decreases the set VM number n (step S54b). ). For example, the VM number determination module 50b decreases the set VM number n by one. Thereafter, the process returns to step S20.

  When the calculated disk bandwidth BW satisfies the guaranteed bandwidth, that is, when the disk bandwidth BW becomes larger than the guaranteed bandwidth (step S53b; Yes), the loop processing ends there. The VM number determination module 50b can determine the maximum VM number n (MBX) satisfying the guaranteed bandwidth based on the set VM number n at this time.

(Step S70b)
The output module 70 b outputs the determined maximum VM number n (MAX) through the output device 5. The user can design the system with reference to the maximum number of VMs n (MAX).

3-3. Effect As described above, according to the third embodiment, the maximum number n (MAX) satisfying the guaranteed bandwidth can be estimated by reducing the set VM number n from the limit value. Thereby, the same effect as the second embodiment is obtained.

  In the second and third embodiments, the increase / decrease width of the VM number n is not limited to 1. Further, the VM number n may be increased or decreased by multiplying the VM number n by a constant or by dividing the VM number n by a constant. Furthermore, it is possible to combine the second embodiment and the third embodiment. For example, when the disk bandwidth BW falls below the guaranteed bandwidth after halving the VM number n from the limit value, the maximum VM number n (MAX) may be obtained by increasing the VM number n by 1.

4). Fourth Embodiment The number of disks is not limited to one and may be plural. For example, a server system that performs enormous request processing such as Web processing may require a plurality of disks in order to ensure disk bandwidth. In this case, it is important to estimate the number of disks sufficient to secure a desired disk bandwidth from the number of virtual machines (virtual servers) operated as servers and the expected request amount.

  The fourth embodiment of the present invention proposes a technique for estimating the number of disks that can guarantee a desired disk bandwidth even when multiple file accesses occur. As will be described below, such a number of disks can also be calculated by applying the first embodiment described above. In the following description, the case of a virtual machine is considered as an example.

4-1. Configuration FIG. 19 is a block diagram showing a configuration of a disk bandwidth calculation system 1c according to the fourth embodiment. The same reference numerals are given to the same components as those in the above-described embodiments, and overlapping descriptions are omitted as appropriate.

  Compared to the first embodiment, the storage device 2 further stores VM size information VMS, VM number information VMN, disk size information DKS, and request information REQ. The VM size information VMS and the disk size information DKS are the same as those in the second and third embodiments. The VM number information VMN indicates the total number of virtual machines assumed in operation.

  The request information REQ is information indicating a request load assumed in operation. FIG. 20A shows an example of the request information REQ. In the example of FIG. 20A, the request information REQ indicates the number of requests per unit time required per virtual machine and the amount of data that needs to be accessed per request. By multiplying the number of requests and the amount of data, the disk bandwidth required for each virtual machine can be calculated. The disk bandwidth required for processing the requested request is hereinafter referred to as “request bandwidth”. The requested bandwidth is a value according to the request load and can be calculated from the request information REQ. In the case of the example shown in FIG. 20A, the required bandwidth per virtual machine is 40.0 [kByte / sec]. FIG. 20B shows another example of the request information REQ. The request information REQ shown in FIG. 20B gives the required bandwidth itself.

  Referring to FIG. 19 again, the processing apparatus 3 includes a data input module 10c, a file position information creation module 20c, a seek distance calculation module 30, a disk bandwidth calculation module 40, a disk number determination module 60, and an output module 70c. is doing. These modules are also realized by the processing device 3 executing the program PROG. Hereinafter, processing by each module will be described in detail.

4-2. Processing Flow FIG. 21 is a block diagram showing a processing flow in the fourth embodiment. FIG. 22 is a flowchart showing the processing in the fourth embodiment. The description overlapping with the above-described embodiment is omitted as appropriate.

(Step S10c)
The data input module 10 c acquires VM size information VMS, VM number information VMN, disk size information DKS, access characteristic information ACS, disk profile information PRF, and request information REQ, and stores them in the storage device 2.

(Step S61)
The disk number determination module 60 is a module that variably sets the disk number p. In step S61, the disk number determination module 60 sets the disk number p to an initial value. The initial value is 1, for example.

  Thereafter, based on the disk number p set by the disk number determination module 60, the file position information creation module 20c, the seek distance calculation module 30, and the disk bandwidth calculation module 40 execute their respective processes. Then, according to the processing result, the disk number determination module 60 changes the disk number p. That is, a loop process is executed for each number of disks p set by the disk number determination module 60.

(Step S20c)
The file position information creation module 20 c creates file position information LOC corresponding to the number of disks p set by the disk number determination module 60. More specifically, the file position information creation module 20c creates the file position information LOC for one disk in consideration of the set number of disks p. For this purpose, the file position information creation module 20 c reads VM size information VMS, VM number information VMN, and disk size information DKS from the storage device 2. Then, the file position information creation module 20c calculates the VM number n per disk by dividing the total number of VMs given by the VM number information VMN by the set disk number p. Thereafter, the file location information creation module 20c creates file location information LOC related to one disk based on the number of VMs n, the VM size, and the like. The creation method is the same as the method described in the second and third embodiments. The created file position information LOC is stored in the storage device 2.

(Step S30)
As in the first embodiment, the seek distance calculation module 30 creates seek distance information SEK based on the file position information LOC created in step S20c.

(Step S40)
Similar to the first embodiment, the disk bandwidth calculation module 40 calculates the disk bandwidth BW when multiple file accesses occur. Any of the above-described methods may be used as the method for calculating the disk bandwidth BW.

(Steps S62 to S64)
The disk number determination module 60 reads VM number information VMN and request information REQ from the storage device 2. Then, the disk number determination module 60 refers to the request information REQ and calculates the required bandwidth per virtual machine. Also, the disk number determination module 60 calculates the VM number n per disk by dividing the total number of VMs given by the VM number information VMN by the current set disk number p. From the required bandwidth per virtual machine and the number of VMs n per disk, the disk number determination module 60 can calculate the required bandwidth RBW per disk. Then, the disk number determination module 60 compares the disk bandwidth BW calculated in step S40 with the requested bandwidth RBW (step S62). Based on the comparison result, the disk number determination module 60 changes the set disk number p.

  Specifically, the disk number determination module 60 determines whether or not the calculated disk bandwidth BW satisfies the required bandwidth RBW. When the disk bandwidth BW does not satisfy the required bandwidth RBW, that is, when the disk bandwidth BW is smaller than the required bandwidth RBW (step S63; No), the disk number determination module 60 increases the set disk number p ( Step S64). For example, the disk number determination module 60 increases the set disk number p by one. Thereafter, the processing returns to step S20c described above. That is, the file position information creation module 20c updates the file position information LOC in accordance with the new set disk count p, the seek distance calculation module 30 creates the seek distance information SEK again, and the disk bandwidth calculation module 40 The bandwidth BW is calculated.

  When the calculated disk bandwidth BW satisfies the required bandwidth RBW, that is, when the disk bandwidth BW becomes larger than the required bandwidth RBW (step S63; Yes), the loop processing ends there. The disk number determination module 60 can determine the minimum necessary number of disks p (min) that satisfies the required bandwidth RBW based on the set number of disks p at this time. This number of disks p (min) is the number of disks necessary for processing the assumed request.

(Step S70c)
The output module 70 c outputs the determined number of disks p (min) through the output device 5. The user can design the system with reference to the number of disks p (min).

4-3. Example of Processing Next, an example of processing will be described using specific numerical values. A VM image file is considered as a file to be accessed. The access characteristic information ACS and the disk profile information PRF are the same as those shown in FIG. The VM size is 10 [GByte]. The disk size is 160 [GByte]. The total number of VMs given by the VM number information VMN is 12. The request information REQ is the same as that shown in FIG. 20A. In this case, the required bandwidth per virtual machine is 40.0 [kByte / sec].

  In step S61, the initial value of the number of disks p is set to 1 (p = 1).

  When p = 1, 12 VM image files are arranged on one disk. Therefore, in step S20, file position information LOC indicating the arrangement positions of the 12 VM image files is created. The method for creating the file position information LOC is the same as the method described in the second embodiment. In step S41, the representative seek distance Lrep is calculated as 43.3 [GByte]. In step S42, the disk bandwidth BW corresponding to the representative seek distance Lrep is calculated as 0.256 [MByte / sec]. On the other hand, since the required bandwidth per virtual machine is 40.0 [kByte / sec], the required bandwidth RBW per disk is 0.480 [MByte / sec]. The calculated disk bandwidth BW is smaller than the required bandwidth RBW (step S63; No). Accordingly, the number of disks p is increased and set to 2.

  In the case of p = 2, 6 VM image files are arranged on one disk. Therefore, in step S20, file position information LOC indicating the arrangement positions of the six VM image files is created. In this case, in step S41, the representative seek distance Lrep is calculated as 23.3 [GByte]. In step S42, the disk bandwidth BW corresponding to the representative seek distance Lrep is calculated as 0.292 [MByte / sec]. On the other hand, since the required bandwidth per virtual machine is 40.0 [kByte / sec], the required bandwidth RBW per disk is 0.240 [MByte / sec]. The calculated disk bandwidth BW is larger than the requested bandwidth RBW (step S63; Yes). That is, the disk bandwidth BW when multiple file accesses occur satisfies the required bandwidth RBW. Accordingly, it is estimated that the number of disks p (min) required for processing the assumed request is “two”.

  In step S70c, the number of disks p (min) = 2 that satisfies the required bandwidth according to the request load is output. In addition, the disk bandwidth per disk BW = 0.292 [MByte / sec] and the disk bandwidth of the entire disk = 0.484 [MByte / sec] when multiple file accesses occur are output. Also good.

4-4. Effect As described above, the disk bandwidth calculation system 1c according to the present embodiment calculates the disk bandwidth BW per disk while changing the set number of disks p. This makes it possible to estimate the number of disks p (min) that satisfies the required bandwidth even when multiple file accesses occur. By considering the estimated number of disks p (min) in the system design in this way, it is possible to prevent a situation in which a request cannot be processed. In the above example, the case of a VM image file has been described, but the same applies to other cases.

5. Others The application of the present invention is not limited to the case of the virtual machine image file under the above-described virtualization environment. The present invention can be similarly applied to a file of a normal OS that is not under a virtual environment. For example, by using the present invention, it is possible to estimate the disk bandwidth when analyzing a large number of log files on a disk. As described above, when the target is not a virtual machine image file, the VM number and the VM size in the foregoing embodiment are replaced with the file number and the file size, respectively.

  Disk access is not limited to read processing. For example, even when the write process is continuously executed for a plurality of files on the disk, the same effect can be obtained by applying the present invention. In general, in write processing, write data is once stored in a buffer and then actually written to the disk during the disk free time. The program that has requested the write process receives a process completion notification when the storage in the buffer is completed, and therefore can execute the next process without waiting for the disk processing time. However, when the program continuously records data on a disk, for example, when passing information of people, things, cars, trains, etc. is recorded in different files according to the type, the amount of data to be written is small. It can increase and the buffer can overflow. In such a case, a write request is issued for a plurality of files on the disk, which is the same as the case of the read processing described so far. By applying the present invention assuming such a situation, it is possible to estimate the disk bandwidth when a write access request is generated for a plurality of files on the disk.

  The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the above-described embodiments, and can be appropriately changed by those skilled in the art without departing from the scope of the invention.

FIG. 1 is a conceptual diagram for explaining a case where a plurality of files on a disk are accessed. FIG. 2 is a block diagram showing the configuration of the disk bandwidth calculation system according to the first embodiment of the present invention. FIG. 3 is a conceptual diagram for explaining file position information and access characteristic information. FIG. 4A is a table showing an example of file position information. FIG. 4B is a table showing another example of file position information. FIG. 5A is a diagram illustrating an example of access characteristic information. FIG. 5B is a table showing another example of access characteristic information. FIG. 6A is a graph showing an example of disk profile information. FIG. 6B is a table showing another example of the disk profile information. FIG. 7 is a block diagram illustrating a processing flow according to the first embodiment. FIG. 8 is a flowchart showing the processing in the first embodiment. FIG. 9 is a conceptual diagram for explaining the processing in the first embodiment. FIG. 10 is a table showing an example of file position information, access characteristic information, and disk profile information. FIG. 11 is a block diagram illustrating a processing flow in the modification. FIG. 12 is a block diagram showing a processing flow in another modification. FIG. 13 is a block diagram showing a configuration of a disk bandwidth calculation system according to the second embodiment of the present invention. FIG. 14 is a block diagram illustrating a processing flow according to the second embodiment. FIG. 15 is a flowchart showing processing in the second embodiment. FIG. 16 is a block diagram showing a configuration of a disk bandwidth calculation system according to the third embodiment of the present invention. FIG. 17 is a block diagram illustrating a processing flow according to the third embodiment. FIG. 18 is a flowchart illustrating the processing according to the third embodiment. FIG. 19 is a block diagram showing a configuration of a disk bandwidth calculation system according to the fourth embodiment of the present invention. FIG. 20A is a table showing an example of request information. FIG. 20B is a table showing another example of request information. FIG. 21 is a block diagram illustrating a processing flow according to the fourth embodiment. FIG. 22 is a flowchart showing the processing in the fourth embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Disk bandwidth calculation system 2 Storage device 3 Processing apparatus 4 Input device 5 Output device 10 Data input module 20 File position information creation module 30 Seek distance calculation module 40 Disk bandwidth calculation module 41 Representative seek distance calculation module 42 Bandwidth calculation module 50 VM number determination module 60 Disk number determination module 70 Output module 100 Disk LOC File position information SEK Seek distance information ACS Access characteristic information PRF Disk profile information VMS VM size information DKS Disk size information GBW Guaranteed bandwidth information VMN VM number information REQ request information L _ii file within the seek distance L _ij between files seek distance Lrep representative seek distance BW disk bandwidth PROG program

Claims (19)

A disk bandwidth calculation system that calculates a disk bandwidth when access occurs to a plurality of files on a disk,
A storage device storing at least disk profile information indicating a statistical relationship between seek distance and disk bandwidth;
Storage device for storing disk profile information indicating a statistical relationship between seek distance and disk bandwidth, and seek distance information indicating one seek distance between different files and within the same file among the plurality of files. When,
A representative seek distance calculation module that calculates a representative seek distance when an access occurs to the plurality of files by referring to at least the seek distance information ;
A disk bandwidth calculation system comprising: a bandwidth calculation module that refers to the disk profile information and calculates a disk bandwidth corresponding to the representative seek distance as a disk bandwidth when the access occurs. The disk bandwidth calculation system according to claim 1 ,
The storage device further stores access characteristic information indicating characteristics of file access in the disk,
The representative seek distance calculation module refers to the access characteristic information, and calculates a weighted average value of the one seek distance indicated by the seek distance information as the representative seek distance. Disk bandwidth calculation system . The disk bandwidth calculation system according to claim 2 ,
The access characteristic information indicates a file switching rate that is a probability that a second file is accessed next to a first file among the plurality of files.
The representative seek distance calculation module is a disk bandwidth calculation system that calculates the weighted average value by using the file switching rate as a weighting coefficient. The disk bandwidth calculation system according to claim 2 ,
The access characteristic information indicates the number of times the same file is accessed continuously,
The representative seek distance calculation module is a disk bandwidth calculation system that calculates the weighted average value by using the number of times. The disk bandwidth calculation system according to claim 1 ,
The representative seek distance calculation module is a disk bandwidth calculation system that calculates a maximum value of the one seek distance indicated by the seek distance information as the representative seek distance. A disk bandwidth calculation system according to any one of claims 1 to 5 ,
In addition, a seek distance calculation module is provided,
The storage device further stores file position information indicating positions of the plurality of files on the disk,
The seek distance calculation module is a disk bandwidth calculation system that creates the seek distance information based on the file position information. The disk bandwidth calculation system according to claim 6 ,
A disk bandwidth calculation system further comprising a file position information creation module that creates the file position information based on the number of the plurality of files and the size of each file. The disk bandwidth calculation system according to claim 7 ,
And a file number determination module for variably setting the number of the plurality of files.
For each set number, the file position information creation module creates the file position information, the seek distance calculation module creates the seek distance information, and the representative seek distance calculation module calculates the representative seek distance. And the bandwidth calculation module calculates the disk bandwidth when the access occurs. The disk bandwidth calculation system according to claim 8 ,
The storage device further stores guaranteed bandwidth information indicating a predetermined guaranteed bandwidth guaranteed for the disk,
The file number determination module compares the calculated disk bandwidth with the guaranteed bandwidth, and changes the set number based on the comparison result to satisfy the guaranteed bandwidth. A disk bandwidth calculation system that calculates the maximum number of files in a disk. The disk bandwidth calculation system according to claim 9 , wherein
The file number determination module increases the setting number when the calculated disk bandwidth is larger than the guaranteed bandwidth after setting the setting number to an initial value,
The disk bandwidth calculation system, wherein the file number determination module calculates the maximum number based on the set value when the calculated disk bandwidth is smaller than the guaranteed bandwidth. The disk bandwidth calculation system according to claim 9 , wherein
The file number determination module calculates a limit value of the set number from the size of the disk and the size of each file,
The file number determination module decreases the setting number when the calculated disk bandwidth is smaller than the guaranteed bandwidth after the setting number is set to the limit value,
The disk bandwidth calculation system, wherein the file number determination module calculates the maximum number based on the set value when the calculated disk bandwidth is larger than the guaranteed bandwidth. The disk bandwidth calculation system according to claim 6 ,
A disk bandwidth calculation system further comprising a file position information creation module that creates the file position information related to one disk based on the number of disks, the number of the plurality of files, and the size of each file. The disk bandwidth calculation system according to claim 12 ,
In addition, a disk number determination module that variably sets the number of disks
For each set number, the file position information creation module creates the file position information, the seek distance calculation module creates the seek distance information, and the representative seek distance calculation module calculates the representative seek distance. And the bandwidth calculation module calculates the disk bandwidth when the access occurs. The disk bandwidth calculation system according to claim 13 ,
The storage device further stores request information indicating a request load,
The disk number determination module compares the calculated disk bandwidth with the bandwidth according to the load, and changes the set number based on the comparison result, so that it is necessary for processing the request. Disk bandwidth calculation system that calculates the number of hard disks. A disk bandwidth calculation system that calculates a disk bandwidth when access occurs to a plurality of files on a disk,
Storage device for storing disk profile information indicating a statistical relationship between seek distance and disk bandwidth, and seek distance information indicating one seek distance between different files and within the same file among the plurality of files. When,
A disk bandwidth calculation system comprising: a disk bandwidth calculation module that calculates a disk bandwidth when an access occurs to the plurality of files with reference to the disk profile information and the seek distance information. A disk bandwidth calculation method for calculating a disk bandwidth when access occurs to a plurality of files on a disk,
(A) Storage device that stores disk profile information indicating a statistical relationship between seek distance and disk bandwidth , and seek distance information indicating a seek distance between different files and within the same file among the plurality of files. Reading from
(B) calculating a representative seek distance when access to the plurality of files occurs by referring to at least the seek distance information ;
(C) referring to the disk profile information, calculating a disk bandwidth corresponding to the representative seek distance as a disk bandwidth when the access occurs, and a disk bandwidth calculation method. A disk bandwidth calculation method for calculating a disk bandwidth when access occurs to a plurality of files on a disk,
(A) Disc profile information indicating a statistical relationship between seek distance and disc bandwidth, and seek distance information indicating a seek distance between different files and within the same file among the plurality of files from the storage device. A reading step;
(B) A disk bandwidth calculation method comprising: referring to the disk profile information and the seek distance information, calculating a disk bandwidth when an access occurs to the plurality of files. A disk bandwidth calculation program for calculating a disk bandwidth when access occurs to a plurality of files on a disk,
(A) Storage device that stores disk profile information indicating a statistical relationship between seek distance and disk bandwidth , and seek distance information indicating a seek distance between different files and within the same file among the plurality of files. Reading from
(B) calculating a representative seek distance when access to the plurality of files occurs by referring to at least the seek distance information ;
(C) Referring to the disk profile information, causing the computer to execute a step of calculating a disk bandwidth corresponding to the representative seek distance as a disk bandwidth when the access occurs. program. A disk bandwidth calculation program for calculating a disk bandwidth when access occurs to a plurality of files on a disk,
(A) Disc profile information indicating a statistical relationship between a seek distance and a disk bandwidth, and seek distance information indicating a seek distance between different files and within the same file among the plurality of files from a storage device. A step of reading;
(B) A disk bandwidth calculation program that causes a computer to execute a step of calculating a disk bandwidth when access to the plurality of files occurs with reference to the disk profile information and the seek distance information.

Images