JP2008046763A - Disk array subsystem and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To enhance the use efficiency of disc devices constituting an array. <P>SOLUTION: A module 125a calculates a presence ratio of I/O (first presence ratio), a presence ratio of HDD I/O (second presence ratio) and a presence ratio of extent chunk I/O (third presence ratio) at the same time by sampling an access request which has not completed as I/O in the middle of execution for extent chunks each belonging to a plurality of extents included in a storage area of a representative HDD of an array for each array. An engine 125c selects a first extent to which an extent chunk whose first presence ratio should be reduced belongs, and a second extent which is included in a representative HDD of an array with a relatively low second presence ratio and to which an extent chunk whose third presence ratio is the lowest belongs. A module 125b executes extent migration for re-arranging data between the first and the second extents. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a disk array subsystem having at least one disk array device that constitutes a logical volume recognized by a host device, and in particular, a virtual storage suitable for autonomously rearranging data in extent units. The present invention relates to a disk array subsystem and a program having a function for creating a disk.

  The disk array subsystem is known as a representative storage system. The disk array subsystem has at least one array (disk array). The array is composed of one or more disk devices, for example, one or more hard disk drives (HDD). Using this array, a virtual storage area recognized by the host device, that is, a logical volume is defined. Access to the disk array subsystem from the host device is performed on this logical volume.

  Here, a disk array subsystem having a plurality of arrays is assumed. In such a disk array subsystem, for example, there may occur a situation in which access is concentrated in one array and almost no access occurs in another array.

  However, in the conventional disk array subsystem, the correspondence between a logical volume and an array (including a disk array device) that constitutes the logical volume is fixed. For this reason, even if the access frequency is uneven for each array, the access frequency for each array cannot be made uniform. Under such circumstances, it is difficult to effectively use the HDD in the disk array subsystem.

  In particular, due to the recent increase in the size of the disk array subsystem and the increase in storage capacity of a single HDD, the frequency of such access bias increases, and this trend is expected to become more prominent in the future. .

Under such a background, for example, Patent Document 1 describes a technique (first prior art) for reconfiguring a logical volume according to the access frequency of a physical volume. Patent Document 2 discloses a technique for detecting the I / O contention frequency of a physical disk constituting a logical disk (logical volume) and relocating a physical disk having a high access frequency among the logical disks to another logical disk. (Second Prior Art) is described.
JP 2004-272324 A Japanese Patent No. 3427763

  According to the first and second prior arts described above, the physical volume (physical disk) is relocated according to the access frequency (I / O contention frequency). That is, the correspondence between a logical volume and an array that constitutes the logical volume is dynamically managed according to the access frequency.

  In recent years, the entire array area (physical storage resource) in the disk array subsystem is divided into areas (extents) of a certain size, and the correspondence between one or more combinations and logical volumes is dynamically managed. The need for a technology (a so-called disk array subsystem virtualization technology) has increased and has been put into practical use. In this virtualization technology, by copying data between extents, it is possible to change the physical location of data while retaining the data.

  Thus, it is conceivable to apply the first or second prior art to the above-described virtualization technology. For example, an access frequency is detected in units of a single HDD area corresponding to extents (hereinafter referred to as extent chunks), and each extent chunk is rearranged across the array (or within the same array) according to the access frequency. It is possible.

However, the method of rearranging each extent chunk according to the access frequency is not necessarily sufficient for the reason described below to effectively use the HDD.
First, the HDD cannot process a plurality of access requests (I / O requests) at the same time. In the HDD, even when a plurality of requests are for different extent chunks, the different extent chunks cannot be accessed simultaneously. Due to such characteristics of the HDD, the HDD has a state in which uncompleted access requests (I / O requests) exist simultaneously for a plurality of extent chunks. Such a situation occurs even if the access frequency of the entire HDD or the access frequency of extent chunks is low. In other words, in consideration of the characteristics of the HDD, the rate at which uncompleted I / O requests for a plurality of extent chunks exist simultaneously is more important than the access frequency (hereinafter referred to as the simultaneous I / O existence rate).

  Since access is concentrated only on a specific extent chunk, an HDD with high access frequency of the entire HDD is assumed. Such an HDD has a low simultaneous I / O existence ratio and a high degree of being in an idle state. For this reason, even if the extent chunks in which accesses are concentrated in the HDD are rearranged, it is not expected to improve the utilization efficiency of the entire HDD.

  Next, even if the access frequency of the entire HDD is high, the frequency with which I / O requests for a plurality of extent chunks are generated at the same time is low, and therefore there are unfinished I / O requests for a plurality of extent chunks at the same time (simultaneous Assume an HDD with a low I / O presence rate. That is, it is assumed that access to the HDD is dispersed over time. In such an HDD, even if extent chunks with high access frequency in the HDD are rearranged, it is not expected to improve the utilization efficiency of the entire HDD.

  The present invention has been made in consideration of the above circumstances, and an object of the present invention is to provide a disk that constitutes an array by performing data relocation between arrays based on the frequency at which I / O requests for a plurality of extent chunks occur simultaneously. It is an object of the present invention to provide a disk array subsystem and program capable of increasing the utilization efficiency of the apparatus.

  According to one aspect of the present invention, the storage area of one or more disk devices includes a plurality of arrays defined as one continuous area, and the entire area of the array is divided into a plurality of extents. By combining the above extents, a disk array subsystem having at least one disk array device constituting a logical volume recognized from the host device is provided. The at least one disk array device has, for each array, an extent chunk belonging to each of the plurality of extents included in a storage area of a predetermined representative disk device among the disk devices constituting the array. A first existence rate representing the degree of simultaneous execution of I / Os in a plurality of extent chunks including the extent chunk by sampling an incomplete access request at the sampling cycle as the executing I / O, the representative Calculating means for calculating a second existence ratio indicating a degree of presence of executing I / O in the disk device and a third existence ratio indicating a degree of existence of executing I / O in the extent chunk; Based on the first to third existence rates, the extent to which the first existence rate should be lowered The second extent is selected for the representative disk device of the array including the first extent, which is different from the representative disk device of the array including the first extent. Selection means for selecting, as a second extent, an extent to which an extent chunk having the lowest third existence ratio belongs from among extent chunks included in a representative disk device of an array having a relatively low existence ratio; Extent migration executing means for executing inter-array extent migration for rearranging data between the first and second extents.

  Here, in order for the calculation means to acquire information serving as an index representing the first existence ratio, a counter (first counter) is provided for each extent chunk of the representative disk device of each array, and a certain period ( (For example, a fixed time period, a random time period, or a period determined by a combination of a fixed time period and a random period) When I / O exists in a plurality of extent chunks at the same time, the counter corresponding to the extent chunk in which the executing I / O exists is incremented by one. In this case, the calculation means obtains the value obtained by dividing the counter value of the counter for each extent chunk by the number of samples, the degree of simultaneous execution of I / O in a plurality of extent chunks including the extent chunk, that is, the first What is necessary is just to acquire as a presence rate.

  In addition, in order for the calculation means to acquire information serving as an index representing the second existence ratio, a counter (second counter) is provided for each representative disk device of each array, and the representative disk device is in a certain cycle. The number of executing I / Os is sampled every time, and the counter corresponding to the representative disk device in which the executing I / O exists is incremented by one. In this case, the calculation means may obtain a value obtained by dividing the count value of the counter for each representative disk device by the number of samples as the degree of presence of the executing I / O in the representative disk device, that is, the second existence rate. good.

  In addition, in order for the calculation means to acquire information serving as an index representing the third existence rate, a counter (third counter) is provided for each extent chunk of the representative disk device of each array, and the calculation unit has a certain period. The number of executing I / Os is sampled for each extent chunk, and the counter corresponding to the extent chunk in which the executing I / O exists is incremented by one. In this case, the calculation means may obtain a value obtained by dividing the count value of the counter for each extent chunk by the number of samples as the degree of execution of I / O in the extent chunk, that is, the third existence ratio.

  Here, the selection means has an extent chunk having the highest first existence rate among extents constituting each of the arrays belonging to an array group predetermined as a data relocation candidate among the plurality of arrays. Of the extent chunks included in the representative disk device having the lowest second existence ratio among the representative disk devices of each of the arrays belonging to the array group. It is preferable that the extent chunk having the lowest third existence rate is selected, and the extent to which the selected extent chunk belongs is selected as the second extent.

Further, as conditions for the selection means to select the first and second extents to be the extent migration target, that is, conditions for executing the next extent migration by the extent migration execution means, the following four It is good to assume that all the conditions are satisfied. These four conditions are
1) Elapsed time since the extent migration was last executed by the extent migration execution means is a predetermined time or more 2) Maximum value and minimum value of the second existence ratio of the representative disk devices of the array belonging to the array group 3) The minimum value of the second existence rate of the representative disk device of the array belonging to the array group is equal to or less than a predetermined value. 4) The extent chunk included in the representative disk device of the array belonging to the array group. The maximum value of the first existence ratio is a certain value or more.

  Here, as a certain time in the above condition, that is, as the extent migration execution cycle, a function of the difference between the maximum value and the minimum value of the second existence rate of the representative disk device of the array belonging to the array group is used. If it is large, the extent migration execution cycle should be shortened.

  Further, when the selection means assumes that N is an integer of 2 or more, the target of any array of the plurality of arrays is the extent chunk included in the representative disk device of the arbitrary array. N extent chunks having a high third existence rate are selected, and one of N extents to which the selected N extent chunks belong or N-1 extents of the N extents is selected as one. Each selected as a third extent, and N or N-1 extents in any of the arrays subject to in-array extent migration with the N or N-1 third extents. And the positions in the array of the N extents to be subject to the extent migration in the array. Alternatively, one of the N extents excluding the N-1 third extent and the position in the array of the N-1 extents to be subject to the extent migration in the array is the array. N or N-1 extents closer to the positions in the array of the N extents to which the N extent chunks having a higher third existence rate belong are selected as fourth extents one by one. The extent migration execution means executes in-array extent migration that rearranges data between the selected third and fourth extents each time the third and fourth extent pairs are selected. You may do it.

  In addition, in order to realize in-array extent migration different from the in-array extent migration described above, the selection unit sets extents constituting an arbitrary array of the plurality of arrays as the third existence rate. The extents to which the highest extent chunk belongs are selected as the third extent one by one in descending order of the third existence rate, and the extents that are subject to in-array extent migration with the selected third extent Are selected as fourth extents one by one from the outer periphery to the inner periphery of the disk device that constitutes the arbitrary array, so that a pair of third and fourth extents to be subjected to the in-array extent migration is sequentially selected. Select the third and fourth extensions Each time the pair of is selected, the extent migration execution means may be configured to perform the array extents migration to relocate data between the third and fourth extent said selected.

  Here, when there are a plurality of extent chunks having the same third existence ratio, the extent to which the extent chunk far from the outer periphery of the representative disk device belongs is set as the third extent first among the plurality of extent chunks. It should be selected.

  As described above, the disk array device may have an intra-array extent migration function in addition to the inter-array extent migration function. Here, the disk array device may have only an in-array extent migration function.

In addition, as a condition for the selection means to select a pair of third and fourth extents to be subjected to the in-array extent migration, that is, as a condition for executing the next in-array extent migration by the extent migration executing means. Suppose that the following two conditions are all satisfied. These two conditions are
1) Elapsed time since the last extent migration in the array was executed by the extent migration execution means for a certain time or more 2) A pair of extent chunks included in a pair of extents that are candidates for the extent migration in the array The difference in the third existence ratio of the pair of extent chunks included in the representative disk device of the array to which the pair of extents belongs is equal to or greater than a certain value.

  Here, in the representative disk device corresponding to a pair of extents that are candidates for in-array extent migration, as a certain period of time during the above-mentioned conditions for executing in-array extent migration, that is, as an execution period of in-array extent migration. The function of the difference between the third existence ratios of the included pair of extent chunks is used, and if the difference is large, the execution period of the in-array extent migration may be shortened.

  Further, in a disk array device having both functions of inter-array extent migration and intra-array extent migration, the extent migration execution means executes extent migration between the arrays and then becomes an object of extent migration between the arrays. The extent migration in the array may be executed for each of the arrays.

  Further, in a disk array subsystem having a plurality of disk array devices connected by a network, after the extent migration execution means of the plurality of disk array devices executes extent migration between the arrays in the disk device, It is preferable that extent migration between the arrays is executed between the plurality of disk array devices, and then extent migration within the array is executed for each of the arrays that are subject to extent migration between the arrays.

  According to the present invention, data is rearranged between arrays on the basis of the degree (first existence rate) at which I / O requests for a plurality of extent chunks in a disk device occur at the same time. The degree of simultaneous I / O requests for a plurality of extent chunks can be reduced to reduce the number of simultaneous I / O requests in the disk device, thereby improving the performance of the disk array subsystem.

Embodiments of the present invention will be described below with reference to the drawings.
FIG. 1 is a block diagram showing a configuration of a disk array subsystem 1 according to an embodiment of the present invention. The disk array subsystem 1 is realized as a storage system (autonomous data relocation storage system) having at least one disk array device, for example, two disk array devices 10-1 and 10-2.

  The disk array devices 10-1 and 10-2 are connected to at least one host device that uses the disk array devices 10-1 and 10-2, for example, the host device 20 via a network 31. The disk array devices 10-1 and 10-2 are also interconnected by a network 32. The networks 31 and 32 are configured by using, for example, Fiber Channel (FC) or Internet SCSI (Internet SCSI).

  The disk array device 10-1 has at least one disk array (array), for example, three arrays 11-1 (# 1), 11-2 (# 2), and 11-3 (# 3). To do. On the other hand, the disk array device 10-2 has at least one disk array (array), for example, three arrays 11-A (#A), 11-B (#B), and 11-C (#C). It shall be.

  Assume that each of the arrays 11-1 to 11-3 and 11-A to 11-C includes at least one HDD (disk device), for example, six HDDs 110. However, in FIG. 1, the arrays 11-1 to 11-3 and 11 -A to 11 -C are shown as being composed of three HDDs 110 for the convenience of drawing. The array 11-i (i = 1, 2, 3, A, B, C) in this embodiment is, for example, RAID (Redundant Array of Independent Disks or Redundant Array of Inexpensive Disks) to which RAID 5 (RAID 5 level) is applied. . The area of the array 11-i is defined as one continuous area by combining the HDDs 110 that constitute the array 11-i and storing the storage areas of these HDDs 110. Each HDD 110 in the disk array devices 10-1 and 10-2 is assigned a unique identification number (HDD number) in the disk array devices 10-1 and 10-2.

  The disk array device 10-1 has an array controller 12-1 that controls access to the arrays 11-1 to 11-3 in response to a request from the host device 20. The disk array device 10-2 has an array controller 12-2 that controls access to the arrays 11-A to 11-C in response to a request from the host device 20. The array controllers 12-1 and 12-2 are connected to the host device 20 via the network 31.

  The array controllers 12-1 and 12-2 manage the entire areas of the arrays 11-1 to 11-3 and 11-A to 11-C, respectively, by dividing (dividing) into physical extents of a certain size. . The array controllers 12-1 and 12-2 constitute a logical volume (logical disk) that is recognized (accessed) from the host device 20 by using one or more physical extents (for example, a plurality of physical extents). That is, the array controllers 12-1 and 12-2 allocate the logical extents constituting the logical volume to the same number of physical extents as the logical extents.

  The association between the logical extent and the physical extent is managed by the mapping table. For example, when the physical extent associated with the logical extent LE1 is changed from the physical extent PE1 to the physical extent PE2, the mapping information indicating the correspondence between LE1 and PE1 in the mapping table is changed to the mapping information indicating the correspondence between LE1 and PE2. Updated. A technique for managing such association between logical extents and physical extents has been well known. For this reason, in the following, description of changing the physical extent associated with the logical extent is omitted. In the following, a physical extent will be described, and a physical extent is simply expressed as an extent unless it is necessary to distinguish between a physical extent and a logical extent.

  Among extents, an area corresponding to (belonging to) each HDD 110 is called an extent chunk. FIG. 2 shows the relationship between extents and extent chunks as an array 11-i (i = 1, 2, 3, A, B, C).

  In this embodiment, the total number of HDDs 110 in each of the disk array devices 10-1 and 10-2 is 18, and the capacity of each HDD 110 is 300 GB. The extent size of each array 11-i is fixed to 1 GB. Each array 11-i is composed of six HDDs 110 as described above (only three HDDs 110 are shown for convenience in the figure), and is a RAID to which RAID 5 is applied.

  The substantial data capacity of each array 11-i is the capacity of five HDDs because the array 11-i applies RAID5. In this case, the extent chunk size is 1 GB / 5 = 0.2 GB. Therefore, the total number of extent chunks per HDD (total number of extents per array) is 300 GB / 0.2 GB = 1500.

  As described above, the array controllers 12-1 and 12-2 map the correspondence between the logical volume accessed from the host device 20 and the logical extent constituting the logical volume, and the logical extent and the logical extent are mapped. Manage the correspondence with physical extents. The array controllers 12-1 and 12-2 perform the physical arrangement of data in the disk array devices 10-1 and 10-2 corresponding to the array controllers 12-1 and 12-2, respectively, or the disk array device. Relocate autonomously across 10-1 and 10-2. More specifically, the array controllers 12-1 and 12-2 perform data copying between the extents, so that the physical extents of the data are stored in the disk array apparatuses 10-1 and 10- while retaining the data. 2 or across the disk array devices 10-1 and 10-2. Such data rearrangement by exchanging extents is called extent migration.

  Here, special terms applied in the present embodiment will be described. In this embodiment, a representative HDD, an executing I / O, a simultaneous I / O existence rate, a HDD I / O existence rate, and an extent chunk I / O existence rate are defined.

<Representative HDD>
The representative HDD indicates a specific one of the six HDDs 110 constituting the array 11-i (i = 1, 2, 3, A, B, C). Here, the HDD with the smallest HDD number among the HDDs 110 constituting the array 11-i is used as the representative HDD.

<Executing I / O>
The in-execution I / O indicates an access request (disk I / O request for read / write) issued to the HDD 110 from the array controller 12-j (j = 1, 2). That is, the in-execution I / O indicates not only an access request for which an access to an extent chunk in the HDD 110 is currently being executed but also an access request that is waiting for an access to an extent chunk. The fact that there is an I / O request (execution I / O, incomplete I / O) that is in the execution wait state for access to the extent chunk or that is being executed, / O exists in an extent chunk "or" an ongoing I / O exists for an extent chunk ". The access request (I / O request) is held as a queue list in the I / O queue (I / O queue buffer) of the HDD 110, and is deleted from the queue when the access specified by the request is completed.

<Simultaneous I / O presence rate>
The simultaneous I / O existence ratio (first existence ratio) indicates the degree of simultaneous execution of I / O in a plurality of extent chunks in the HDD 110 (representative HDD 110). This simultaneous I / O existence rate is calculated for each extent chunk. In other words, the simultaneous I / O existence ratio indicates the degree of simultaneous execution of I / O in the extent chunk of interest and other extent chunks in the HDD 110 (representative HDD 110). In general, the arrays 11-1 to 11-3 and 11-A to 11-C are often accessed equally to each RAID chunk unit, that is, each extent chunk constituting the extent. Therefore, the simultaneous I / O existence rate of the extent chunk in the representative HDD 110 can be regarded as representing the simultaneous I / O existence rate of the extent to which the extent chunk belongs. Further, for example, in the case of RAID1, when the read data is read from one of the paired HDDs, the HDD reflecting the I / O state of the array is selected as the representative HDD, such as the HDD that reads the read data as the representative HDD. And good.

<Extent chunk I / O existence rate>
The extent chunk I / O existence rate (third existence rate) is calculated for each extent chunk in the representative HDD 110, and there is an executing I / O (at least one executing I / O) in the extent chunk. Refers to degree.

<HDD I / O presence rate>
The HDD I / O existence rate (second existence rate) indicates the degree of execution of I / O being executed (at least one executing I / O) in the representative HDD 110 calculated for each representative HDD 110.

  The array controllers 12-1 and 12-2 perform the extent migration described above between the first extent of the first array and the second extent of the second array described below.

  The first extent is a first array to which an extent chunk having a simultaneous I / O existence rate in a representative HDD 110 (first representative HDD 110) constituting a certain array (first array) higher than a certain level belongs. Points to the extent within The second array is an array different from the first array, and the degree to which the executing I / O is present in the representative HDD (second representative HDD 110) of the second array (that is, the HDD I / O exists) (Rate) refers to an array that is below a certain level. The second extent is an extent chunk (that is, extent chunk I / O) that has the lowest rate of presence of executing I / O in the second representative HDD 110 in the second array among the extents in the second array. This refers to the extent to which the extent chunk having the lowest O existence rate belongs.

  In this embodiment, the array controllers 12-1 and 12-2 have the same configuration. Therefore, the configuration of the array controllers 12-1 and 12-2 will be described by using the array controller 12-j (j = 1, 2) as a representative.

  FIG. 3 is a block diagram showing the configuration of the array controller 12-j. If necessary, the array controller 12-j should be appropriately replaced with the array controller 12-1 or 12-2 in the following description.

  The array controller 12-j includes a host driver 121, a host manager 122, a cache 123, a cache controller 124, a RAID manager 125, an HDD driver 126, and a memory 127.

  The host driver 121 inputs and outputs data commands with the host device 20. The host manager 122 manages an array (arrays 11-1 to 11-3 or arrays 11-A to 11-C in FIG. 1) that provides a logical volume recognized by the host device 20. The cache 123 temporarily stores data input / output between the host device 20 and the array. The cache controller 124 controls access to the cache 123.

  The RAID manager 125 manages the correspondence between arrays that provide logical volumes recognized by the host device 20 and extents, and the correspondence between extents and extent chunks in the HDD 110. The RAID manager 125 has a function of calculating the simultaneous I / O presence rate, the HDD I / O presence rate, and the extent chunk I / O presence rate, and a function of executing the extent migration described above based on these presence rates.

  The HDD driver 126 executes access (disk I / O) to each HDD 110 constituting the array. The HDD driver 125 measures the number of HDD I / Os N1 for each representative HDD 110 and the number of extent chunk I / Os N2 for each extent chunk in the representative HDD 110. The HDD I / O number N1 represents the number of I / Os being executed in the representative HDD 110, that is, the number of incomplete I / Os present in the representative HDD 110. The number N2 of extent chunk I / Os represents the number of I / Os being executed in the extent chunks, that is, the number of incomplete I / Os present in the extent chunks.

  The memory 127 holds various counters (count values thereof) including the counters (N1 counter and N2 counter) for measuring N1 and N2 described above. These counters are software counters.

  In this embodiment, the host driver 121, the host manager 122, the cache controller 124, the RAID manager 125, and the HDD driver 126 are realized by the array controller 12-j reading and executing a predetermined program from a storage medium. This storage medium is, for example, a magnetic disk medium mounted on an HDD, an optical disk medium used by being mounted on an optical disk drive, or a ROM. Further, the above program may be downloaded to the array controller 12-j via the network 31, for example.

Next, for example, extent migration by the array controller 12-1 (internal RAID manager 125) will be described with reference to FIG.
In FIG. 4, the representative HDD 110 (# 1) represents the representative HDD 110 of the first array 11-1 (# 1). In each of the extent chunks 41 to 43 in the representative HDD 110 (# 1), the degree (simultaneous I / O existence rate) that a plurality of extent chunks including the extent chunk are simultaneously executing I / O (simultaneous I / O existence rate) is from a certain level. Are also high extent chunks. That is, the extent chunks 41 to 43 are assumed to be a set of extent chunks having a high simultaneous I / O existence rate. Here, extent chunks that have a high degree of the number of executing I / Os being two or more are treated as extent chunks that have a high simultaneous I / O existence rate. Such extent chunks do not respond well.

  On the other hand, the representative HDD 110 (# 2) represents a second HDD different from the array 11-1 (# 1), for example, the representative HDD 110 of the array 11-2 (# 2). The representative HDD 110 (# 2) has a lower degree of presence of executing I / O (HDD I / O presence rate) than the representative HDD 110 (# 1), and therefore has a lower HDD usage rate (utilization rate). . The extent chunk 44 in the representative HDD 110 (# 2) has the lowest degree of I / O being executed (extent chunk I / O existence rate) among the extent chunks in the representative HDD 110 (# 2). To do.

  The array controller 12-1 (internal RAID manager 125) is one of the extent chunks 41 to 43, for example, the extent (first extent) in the array 11-1 (# 1) to which the extent chunk 41 belongs, Extent migration (inter-array extent migration) 45 is performed between the extent (second extent) in the array 11-2 (# 2) to which the extent chunk 44 belongs.

  By this extent migration 45, the data D1 of the extent chunk 41 (with a high simultaneous I / O existence rate) in the representative HDD 110 (# 1) is stored in the representative HDD 110 (# 2) (with a low HDD I / O existence rate) ( The data D2 of the extent chunk 44 is rearranged in the extent chunk 41 in the extent chunk 44 (with the lowest extent chunk I / O existence ratio). In this extent migration 45, the data of other extent chunks included in the extent (first extent) to which the extent chunk 41 belongs is stored in other extent chunks included in the extent (second extent) to which the extent chunk 44 belongs. The data of another extent chunk included in the second extent is relocated to the other extent chunk included in the first extent.

  That is, in the extent migration 45, the data of the first extent described above is rearranged in the second extent, and the data of the second extent is rearranged in the first extent. Here, it is assumed that the first and second extents (physical extents) are associated with the first and second logical extents before the extent migration 45 is executed. In this case, as the extent migration 45 is executed, the above-described mapping table is updated, and the first and second extents (physical extents) are associated with the second and first logical extents, respectively. Since the mapping table update (mapping change) itself associated with the execution of extent migration is well known in the past, it will be omitted in the following description of the new extent migration.

  As a result of the data relocation 46 by the extent migration 45, the degree to which access requests (I / O requests) are simultaneously generated in a plurality of extent chunks in the HDD 110 constituting the array 11-1 (# 1) is lowered. That is, the number of executing I / Os simultaneously existing in the HDDs 110 constituting the array 11-1 (# 1), that is, executing I / Os connected to the HDD I / O queue (incomplete I / O requests). Number) is reduced. As a result, the utilization efficiency (operation rate) of the HDDs 110 constituting the array 11-1 (# 1) can be increased, and the performance of the HDDs 110 can be utilized as much as possible.

Hereinafter, the extent migration described above will be described more specifically.
In the example of FIG. 4, a set of extent chunks 41 to 43 having a high simultaneous I / O existence rate exists in the HDD 110 (representative HDD 110 (# 1)) configuring the array 11-1 (# 1). In such an HDD 110, the number of simultaneous I / Os (the number of executing I / Os connected to the HDD I / O queue) is likely to be 2 or more. Since the HDD 110 cannot process a plurality of I / O requests at the same time, the response tends to decrease when the number of simultaneous I / Os is 2 or more.

  On the other hand, the HDD I / O existence rate of the HDD 110 (representative HDD 110 (# 2)) constituting the array 11-2 (# 2) is low. When such an array 11-2 (# 2) exists, the HDD 110 constituting the array 11-2 (# 2) has a high ratio of being in an idle state. That is, the performance of the HDD 110 that constitutes the array 11-2 (# 2) cannot be fully utilized.

  In this embodiment, the extent migration 45 between the two arrays 11-1 (# 1) and 11-2 (# 2) described above causes the simultaneous I / O existence rate in the array 11-1 (# 1). One extent included in the extent group having a high level and the extents in the other array 11-2 (# 2) (that is, the array 11-2 (# 2) configured by the HDD 110 having a low HDD I / O existence rate) Data is exchanged (rearranged) with the extent to which the extent chunk with the lowest chunk I / O existence rate belongs. This makes it possible to equalize the simultaneous I / O existence rate in each HDD 110 in the disk array device 10-1 and to utilize the performance of the HDD 110 as much as possible.

  Next, the reason why the concurrent I / O existence rate is used instead of the extent chunk access frequency to determine the extents to be subject to extent migration will be described with reference to FIG. FIG. 5 shows an access pattern example of the HDD 110 for which performance improvement cannot be expected by extent migration for the representative HDD 110 of the array 11-i.

  The example in FIG. 5A assumes a case where accesses (access requests) are concentrated only on one extent chunk 51 in the representative HDD 110. That is, it is assumed that accesses are concentrated on only one extent in the array 11-i.

  In the representative HDD 110 in FIG. 5A, access to extent chunks other than the extent chunk 51 hardly occurs. That is, the representative HDD 110 in FIG. 5A has a high degree of idling even if the access frequency is high due to the concentration of access to only the extent chunk 51. For this reason, even if the extent (data) to which the extent chunk 51 belongs is rearranged, although the access frequency of the HDD 110 constituting the array 11-i can be lowered, it cannot be expected to improve the utilization efficiency.

  In the example of FIG. 5B, it is assumed that in the representative HDD 110, access frequency is high, but access to a plurality of extent chunks (here, extent chunks 52 and 53) is rare. That is, it is assumed that the array 11-i is less frequently accessed at the same time even if the access frequency is high and the load is high. As an example of this, there is a case where the extent to which the extent chunk 52 belongs and the extent to which the extent chunk 53 belongs are alternately accessed.

  In the representative HDD 110 shown in FIG. 5B, accesses to the extent chunks 52 and 53 are temporally dispersed. Therefore, even if the extent to which the extent chunk 52 or 53 belongs is rearranged (distributed), although the access frequency of the HDD 110 constituting the array 11-i can be lowered, it cannot be expected to improve the utilization efficiency.

  In consideration of the above points, in this embodiment, the simultaneous I / O existence rate related to the extent chunk of the representative HDD is used to determine the high-load array to be subject to extent migration and the extent (first extent) in the array. Is used.

  In this embodiment, the extent chunk I / O in the representative HDD of another array (second array) with a low HDD I / O existence rate is used as the extent (second extent) in the array that is the extent migration partner. The extent to which the extent chunk with the lowest O existence degree belongs is used. Two reasons why such an extent is used as a low-load extent of an array that is an extent migration partner will be described.

  First, the first reason will be described. The HDD 110 (representative HDD 110) with a low HDD I / O presence rate has a high idle rate. For this reason, by selectively using such an array constituted by the HDDs 110 as a second array to be an extent migration partner (underwriting destination) with an extent having a high simultaneous I / O existence rate, The operating rate of the HDD 110 can be increased and the performance of the HDD 110 can be utilized.

  Next, the second reason will be described. First, out of the extents in the second array selected based on the first reason, the extent to which the extent chunk having the lowest extent chunk I / O existence rate in the representative HDD belongs is the first in the first array. It is assumed that it is used as a second extent (low load extent) that is a counterpart of extent migration with one extent. In this case, it is possible to prevent as much as possible the degree of I / O being executed in the first array (the array whose load is to be reduced) after the extent migration has been executed.

Next, in the present embodiment, refer to FIG. 6 for a method of acquiring information serving as an index indicating the degree of simultaneous execution of I / Os in a plurality of extent chunks in the representative HDD 110 (simultaneous I / O presence rate). To explain.
6A, the storage area of the representative HDD 110 of the array 11-i is managed by being divided into a plurality of extent chunks 0 (EC-0), 1 (EC-1), 2 (EC-2),. Indicates that

  FIG. 6B shows an example of the number of executing I / Os (that is, the number of in-flight executing I / Os) existing in each extent chunk of the representative HDD 110 sampled every certain sampling period (sampling period). . FIG. 6B also shows counters (simultaneous I / O existence rate calculation counter, first counter) 61-0, 61 used to calculate the simultaneous I / O existence rate for each extent chunk of the representative HDD 110. .., 61-2,... Are shown in association with the simultaneous I / O existence rate (X). As the sample period, a fixed time period, a random time period, or a period determined by a combination of the fixed time period and the random period can be used.

  The counters 61-0, 61-1, 61-2,... Are stored in the memory 127 of the array controller 12-j (j = 1, 2) for each extent chunk of the representative HDD 110 of each array 11-i, that is, extents. These are provided in correspondence with chunks 0 (EC-0), 1 (EC-1), 2 (EC-2),.

  For each extent chunk in the representative HDD 110, the array controller 12-j (internal HDD driver 126) measures the number of executing I / Os N2 existing in the extent chunk using the N2 counter. Here, the number of executing I / Os N2 for each extent chunk (the count value of the N2 counter) is incremented by 1 (+1) every time an I / O request for the extent chunk is issued to the representative HDD 110, and the I / O concerned Each time the execution of the request is completed, 1 is decremented (−1).

  The array controller 12-j (inside the RAID manager 125), for each extent chunk in the representative HDD 110, in a certain sample period, the number of I / Os being executed in the extent chunk (the N2 counter held in the memory 127) (Count value) is sampled. As a result of this sampling, the array controller 12-j (inside the RAID manager 125) converts a plurality of extent chunks including extent chunks 0 (EC-0), 1 (EC-1), 2 (EC-2),. When it is detected that I / Os being executed exist simultaneously, the counters 61-0, 61-1, 61-2, ... are incremented by 1 (+1).

  In the example of FIG. 6, in the first sample period (first sample), I is being executed in extent chunks 1 (EC-1), 2 (EC-2), 3 (EC-3), and 4 (EC-4). As a result of the presence of / O, the counters 61-1, 61-2, 61-3, and 61-4 are incremented by 1 as indicated by the arrow 62, and all of the count values are 1 are shown. Yes.

  Similarly, in the example of FIG. 6, as a result of the execution of I / O in extent chunks 2 (EC-2) and 3 (EC-3) in the third sample period, as indicated by the arrow 63, the counter 61 -2 and 61-3 are incremented by 1, and the count value is 2 in both cases. Similarly, in the example of FIG. 6, as a result of the execution of I / O in extent chunks 1 (EC-1) and 2 (EC-2) in the fourth sample period, as indicated by the arrow 64, the counter 61 -1 and 61-2 are incremented by 1 and the count values become 2 and 3, respectively.

  The array controller 12-j (internal RAID manager 125) includes counters 61-0 and 61 corresponding to extent chunks 1 (EC-1), 2 (EC-2), 3 (EC-3),. −1, 61-2,... Divided by the number of samples (number of samplings), the extent chunks 1 (EC-1), 2 (EC-2), 3 (EC-3),. Acquired as the simultaneous I / O presence rate. In the example of FIG. 6 (example of sampling five times), the extent chunk with the highest simultaneous I / O existence rate is extent chunk 2 (EC-2), that is, extent chunk of extent 2.

  Next, in the present embodiment, a method for acquiring information serving as an index indicating the degree of presence of executing I / O in the representative HDD 110 of the array 11-i (HDD I / O presence rate) will be described with reference to FIG. To do. Here, as shown in FIG. 6A, the storage areas of the representative HDD 110 of the array 11-i are extent chunks 0 (EC-0), 1 (EC-1), 2 (EC-2), It shall be managed by dividing into….

  FIG. 7 shows an example of the number of executing I / Os present in each extent chunk of the representative HDD 110 sampled at a certain sampling period, and a change example of the count value of the I / O presence counter (second counter) 70. Shown in correspondence. The I / O presence counter 70 is provided for each representative HDD 110 of each array 11-i in the memory 127 of the array controller 12-j (j = 1, 2).

  The array controller 12-j (inside the RAID manager 125) samples the number of I / Os in execution existing in the extent chunk for each extent chunk in the representative HDD 110 at a certain sampling period. As a result of this sampling, the array controller 12-j (internal RAID manager 125) is executing one of extent chunks 0 (EC-0), 1 (EC-1), 2 (EC-2),. When it is detected that / O exists, that is, when it is detected that at least one executing I / O exists in the representative HDD 110, the I / O presence counter 70 corresponding to the representative HDD 110 is incremented by 1 ( +1). In the example of FIG. 7, it is detected that there is at least one in-execution I / O in the representative HDD 110 in three periods of the 0th sample period, the first sample period, and the third sample period. The O presence counter 70 is incremented by one.

  The array controller 12-j (within the RAID manager 125) determines the degree to which the I / O being executed exists in the representative HDD 110 by dividing the value obtained by dividing the count value of the I / O presence counter 70 by the number of samples (number of samplings) (HDDI / O abundance ratio). In the example of FIG. 7 (example of sampling five times), 3/5 is calculated as the HDD I / O existence ratio of the representative HDD 110.

  Next, in the present embodiment, information serving as an index indicating the degree (extent chunk I / O existence ratio) of executing I / O in each extent chunk included in the representative HDD 110 of the array 11-i is acquired. The method will be described with reference to FIG. Here, as shown in FIG. 6A, the storage areas of the representative HDD 110 of the array 11-i are extent chunks 0 (EC-0), 1 (EC-1), 2 (EC-2), It shall be managed by dividing into….

  FIG. 8 shows an example of the number of executing I / Os present in each extent chunk of the representative HDD 110 that is sampled every certain sampling period. FIG. 8 also shows counters (extent chunk I / O existence ratio calculation counter, third counter) 81-0, 81- used to calculate the extent chunk I / O existence ratio for each extent chunk of the representative HDD 110. An example of a change in the count value of 1, 81-2,... Is shown in association with the extent chunk I / O existence rate (Y).

  The counters 81-0, 81-1, 81-2,... Are stored in the memory 127 of the array controller 12-j (j = 1, 2) for each extent chunk of the representative HDD 110 of each array 11-i, that is, extents. These are provided in correspondence with chunks 0 (EC-0), 1 (EC-1), 2 (EC-2),.

  The array controller 12-j (inside the RAID manager 125), for each extent chunk in the representative HDD 110, in a certain sample period, the number of I / Os being executed in the extent chunk (the N2 counter held in the memory 127) (Count value) is sampled. As a result of this sampling, the array controller 12-j (the RAID manager 125 in the array controller 12-j) is executing I / O on extent chunks 0 (EC-0), 1 (EC-1), 2 (EC-2),. When it is detected that there is at least one executing I / O, the counters 81-0, 81-1, 81-2, ... are incremented by 1 (+1).

  In the example of FIG. 8, as a result of the execution of I / O in extent chunk 0 (EC-0) in the 0th sample period, the counter 81-0 is incremented by 1 as indicated by the arrow 82, and the count value It is shown that became 1.

  Similarly, in the example of FIG. 8, in the first sample period, extent chunks 1 (EC-1), 2 (EC-2), 3 (EC-3), and 4 (EC-4) are executing I / Os. As a result, as shown by an arrow 83, the counters 81-1, 81-2, 81-3 and 81-4 are incremented by 1, and the count value is 1 in all cases. In the example of FIG. 8, as a result of the execution of I / O in extent chunk 1 (EC-1) in the second sample period, the counter 81-1 is incremented by 1 as indicated by the arrow 84, and the count is performed. A state in which the value is 2 is shown.

  Similarly, in the example of FIG. 8, as a result of the execution of I / O in extent chunks 2 (EC-2) and 3 (EC-3) in the fourth sample period, 8-2 and 81-3 are incremented by 1, and the state in which the count value is 2 is shown. In the example of FIG. 8, as a result of the execution of I / O in extent chunks 1 (EC-1) and 2 (EC-2) in the fifth sample period, as shown by the arrow 86, the counter 81- 1 and 81-2 are incremented by 1, and a state in which the count value is 3 is shown.

  The array controller 12-j (within the RAID manager 125) has counters 81-0 and 81 corresponding to extent chunks 1 (EC-1), 2 (EC-2), 3 (EC-3),. −1, 81-2,... Divided by the number of samples (number of samplings), the extent chunks 1 (EC-1), 2 (EC-2), 3 (EC-3),. This is acquired as the degree of existence of an executing I / O (that is, the extent chunk I / O existence rate) (Y).

Next, conditions for executing extent migration (extent rearrangement) between arrays applied in the present embodiment will be described.
In this embodiment, the array controller 12-j (inside the RAID manager 125) has the following four conditions: 1) The elapsed time since the last extent rearrangement is longer than a certain time 2) the disk array device 10-j The difference between the maximum value and the minimum value among the HDD I / O presence rates of the representative HDDs 110 of all the arrays belonging to an array group (RAID group) that is predetermined as an extent rearrangement candidate (optimization candidate) is constant. 3) The minimum value of the HDD I / O existence rate of the representative HDD 110 of all the arrays belonging to the RAID group is below a certain value. 4) Simultaneously executing on a plurality of extent chunks of the representative HDD 110 of all the arrays belonging to the RAID group I When the maximum value of the degree of / O exists (simultaneous I / O presence rate) exceeds a certain value Run the extents relocated to.

  The reason for setting the above conditions is as follows. First, extent migration involves HDD I / O inside the disk array subsystem (storage system). For this reason, the performance of the disk array subsystem may be reduced while the extent migration is executed. In order to prevent the performance degradation of the disk array subsystem as much as possible, unnecessary extent migration may be avoided. Therefore, in the present embodiment, unnecessary extent migration that is not expected to be very effective is avoided under the conditions 1) to 4). Thereby, stable autonomous data rearrangement becomes possible.

  Here, the execution period T according to the difference Δ between the maximum value and the minimum value of the HDD I / O existence rate of the representative HDD 110 of the array belonging to the RAID group, as the “certain time” in the condition 1), that is, the extent relocation execution cycle. If the difference is larger than a certain level, the extent rearrangement cycle may be shortened. In this way, performance can be improved efficiently. As the function f (Δ), for example, when Δ is less than a certain level, f (Δ) indicates the first execution cycle T1, and when Δ is greater than or equal to a certain level, f (Δ) is shorter than the first execution cycle T1. What shows the execution cycle T2 may be used. It is also possible to use a function f (Δ) that is determined such that when Δ is greater than or equal to a certain level, the Δ becomes shorter as Δ increases.

  Next, details of the calculation of the above-described three types of I / O existence ratios, that is, HDD I / O existence ratio, extent chunk I / O existence ratio, and simultaneous I / O existence ratio will be described. As described above, the HDD I / O presence rate, extent chunk I / O presence rate, and simultaneous I / O presence rate are the I / O presence counter 70 and the extent chunk I / O presence rate calculation counters 81-0 and 81, respectively. .., 81-2,... And the simultaneous I / O existence rate calculation counters 61-0, 61-1, 61-2,.

  The count value of the counter 70, that is, the count value of the counter 70 provided for each representative HDD 110 of each array 11-i is represented by P [AID]. The AID is an identification number (hereinafter referred to as array management) for managing the array 11-i (i = 1, 2, 3 in the disk array device 10-1 and i = A, B, C in the disk array device 10-2). Number). In the present embodiment, a serial number starting from 0 is used as the array management number AID. The count value P [AID] represents the number of samples (sample period) in which incomplete I / O (execution I / O) is present in the representative HDD 110 of the array whose array management number is AID (the number of HDD I / O existing samples). .

  The count values of the counters 81-0, 81-1, 81-2,..., That is, extent chunks 0 (EC-0), 1 (EC-1), 2 (EC-2) of the representative HDD 110 of each array 11-i. ,..., The count values of counters 81-0, 81-1, 81-2,... Are represented by Q [AID] [EID]. The EID is an identification number (hereinafter referred to as an in-array extent management number) for managing extents constituting the array 11-i. In this embodiment, a serial number starting from 0 is used for the extent management number EID in the array. The count value Q [AID] [EID] is the number of samples (extents) in which I / O exists in extent chunks in the representative HDD 110 that belong to an extent whose extent management number in the array that constitutes the array whose array management number is AID. Represents the number of chunk I / O existing samples).

  .., That is, extent chunks 0 (EC-0), 1 (EC-1), 2 (EC-2) of the representative HDD 110 of each array 11-i. ,..., The count values of counters 61-0, 61-1, 61-2,. The count value R [AID] [EID] is an I / O to a plurality of extent chunks including extent chunks in the representative HDD 110 that belong to an extent whose in-array extent management number constitutes an array whose array management number is AID. Represents the number of samples (number of simultaneous I / O existing samples).

  Next, the HDD I / O existence sample number P [AID], extent chunk I / O existence sample, which are used to calculate the HDD I / O existence ratio, extent chunk I / O existence ratio, and simultaneous I / O existence ratio, respectively. Details of collection of the number Q [AID] [EID] and the number of simultaneous I / O existing samples R [AID] [EID] will be described.

  FIG. 9 is a block diagram showing the main configuration of the RAID manager 125 in the array controller 12-j (j = 1, 2) in association with other elements in the array controller 12-j. The RAID manager 125 includes an I / O statistical information collection module 125a, an extent migration execution module 125b, and an autonomous optimization control engine 125c.

  The I / O statistical information collection module 125a samples the HDD I / O count N1 for each representative HDD 110 and the extent chunk I / O count N2 for each extent chunk in the representative HDD 110, which are measured by the HDD driver 126, at a sampling period. . As described above, the HDD I / O number N1 represents the number of incomplete I / Os present in the representative HDD 110. The number of extent chunk I / O N2 represents the number of incomplete I / Os present in extent chunks in the representative HDD 110. The I / O statistical information collection module 125a uses the I / O existence counter 70 based on the sampling result of the HDD I / O count N1 for each representative HDD 110 to obtain a sample (sample) in which the representative HDD 110 has incomplete I / O. Cycle) number (number of HDDI / O existing samples) P [AID] is counted.

  The I / O statistical information collection module 125a also uses extent chunk I / O existence rate calculation counters 81-0, 81-1, 81-2,... Based on the number N2 of extent chunk I / Os for each extent chunk. Thus, the number of samples in which incomplete I / O is present in the extent chunk (number of extent chunk I / O existing samples) Q [AID] [EID] is counted. Further, the I / O statistical information collection module 125a further determines the simultaneous I / O existence ratio counters 61-0 and 61 based on the count result of the number of extent chunk I / O existing samples Q [AID] [EID] for each extent chunk. -1, 61-2,..., For each extent chunk in the representative HDD 110, the number of samples in which I / O is present simultaneously in a plurality of extent chunks including the extent chunk (number of simultaneous I / O existing samples) R [AID] [EID] is counted.

  The I / O statistical information collection module 125a functions as a calculation unit, and calculates the HDD I / O existence rate Pr [AID] based on the HDD I / O existence sample number P [AID]. The I / O statistical information collection module 125a also functions as a calculation unit, and calculates the extent chunk I / O existence rate Qr [AID] [EID] based on the number of extent chunk I / O existence samples Q [AID] [EID]. calculate. The I / O statistical information collection module 125a further functions as a calculation unit, and calculates the simultaneous I / O presence rate Rr [AID] [EID] based on the number of simultaneous I / O existing samples R [AID] [EID]. .

  The extent migration execution module 125b executes extent migration. The autonomous optimization control engine 125c functions as a selection unit, and includes an HDD I / O existence rate Pr [AID] and an extent chunk I / O existence rate Qr [AID] [EID] calculated by the I / O statistical information collection module 125a. Based on the concurrent I / O existence rate Rr [AID] [EID], the extents (first and second extents) that are the targets of extent migration (here, extent migration between arrays) are selected. The autonomous optimization control engine 125c controls the execution of the migration (extent migration) by the extent migration execution module 125b between the selected first and second extents, thereby optimizing the RAID group (autonomous optimization). Control for.

  Next, the HDD I / O existence sample number P [AID], extent chunk I / O existence sample number Q [AID] [EID], and simultaneous I / O existence sample number executed by the I / O statistical information collection module 125a. A procedure in one sample (sample period) of the process of collecting R [AID] [EID] will be described with reference to the flowchart of FIG.

  First, the module 125a performs initialization processing prior to processing for each sample period. In this initialization process, if the maximum value of AID is MaxAID, all P [AID] (counter 70) of AID = 0 to MaxAID are set to the initial value 0. If the maximum value of EID is MaxEID, Q [AID] [EID] (counters 81-0, 81-1,...) Of all combinations of AID = 0 to MaxAID and EID = 0 to MaxEID, and AID = R [AID] [EID] (counters 61-0, 61-1,...) Of all combinations of 0 to MaxAID and EID = 0 to MaxEID are set to an initial value 0.

  The module 125a sets the array management number AID to the initial value 0 for each sample period, for example, at the timing when the end of each sample period arrives (step S1). The module 125a samples the incomplete I / O number (HDD I / O number) N1 of the current sample period in the representative HDD 110 of the array whose array management number is AID, which is measured by the HDD driver 126 (step S2).

  Next, the module 125a determines whether or not the number of sampled HDD I / Os N1 is greater than 0 (step S3). If the HDD I / O number N1 is greater than 0, the module 125a increments P [AID] by 1 (step S4). The module 125a sets the in-array extent management number EID to the initial value 0 (step S5), and sets the temporary counter CNT to the initial value 0 (step S6). The temporary counter CNT is provided in the memory 127.

  The module 125a is an incomplete I / O in the current sample period corresponding to an extent chunk belonging to an extent whose in-array extent management number is EID in the representative HDD 110 of the array whose array management number is AID, which is measured by the HDD driver 126. The number (extent chunk I / O number) N2 is sampled (step S7). In the following description, in order to avoid complicated expression, it is expressed as an EID extent chunk or an extent chunk indicated by EID.

  Next, the module 125a determines whether or not the number N2 of sampled extent chunk I / Os is greater than 0 (step S8). If the extent chunk I / O number N2 is greater than 0, the module 125a increments Q [AID] [EID] by 1 (step S9) and increments CNT by 1 (step S10).

  Next, the module 125a determines whether the incremented CNT value is greater than 1 (step S11). This determination is equivalent to checking whether the number N2 of extent chunk I / Os is determined to be larger than 0 in a plurality of extent chunks in the current array indicated by AID.

  When the value of CNT is not greater than 1 (step S11), that is, when the value of CNT is 1, the module 125a determines that the extent chunk of the current array indicated by AID and EID is N2> 0 in the current array. Recognize that it is the first extent chunk determined. In this case, the module 125a stores the current EID in the memory 127 as a special EID (hereinafter referred to as EIDx) (step S12). Then, the module 125a increments EID by 1 (step S13).

  On the other hand, if the CNT value is greater than 1 (step S11), the module 125a determines whether the CNT value is 2 (step S14). If the value of CNT is 2, the module 125a indicates that the extent chunk of the current array indicated by AID and EID is the extent chunk that is secondly determined as N2> 0 in the current array. recognize. In this case, the module 125a increments R [AID] [EIDx] specified by the current AID and the EIDx stored in the memory 127 in the previous step S12 (step S15). At this time, the module 125a deletes EIDx from the memory 127. Further, the module 125a increments R [AID] [EID] by 1 (step S16). Then, the module 125a increments EID by 1 (step S13).

  On the other hand, if the value of CNT is not 2 (step S14), that is, if the value of CNT is greater than 2, module 125a indicates that the extent chunk of the current array indicated by AID and EID is N2> 0 in the current array. And the third and subsequent extent chunks are recognized. In this case, the module 125a increments R [AID] [EID] by 1 (step S16). Then, the module 125a increments EID by 1 (step S13).

  When executing step S13, the module 125a determines whether the incremented EID is equal to or less than MaxEID (step S17). If EID is less than or equal to MaxEID, the module 125a returns to step S7. On the other hand, if EID is not less than MaxEID, that is, if EID exceeds MaxEID, module 125a determines that processing has been completed for all extents in the array indicated by AID. In this case, the module 125a increments AID by 1 (step S18).

  If the sampled HDD I / O number N1 is not greater than 0 (step S3), that is, if N1 = 0, the module 125a proceeds from step S3 to step S18 and increments AID by 1.

  When executing step S18, the module 125a determines whether the incremented AID is equal to or less than MaxAID (step S19). If the AID is equal to or less than MaxAID, the module 125a returns to step S2. On the other hand, if the AID is not equal to or less than MaxAID, that is, if the AID exceeds MaxAID, the module 125a ends the collection process in the current sample period.

  The module 125a collects the P [AID], Q [AID] [EID] and R [AID] [EID] collected as described above, and the number of samples (number of sample periods) used to collect them. Divide by. As a result, the module 125a acquires the HDD I / O presence rate Pr [AID], the extent chunk I / O presence rate Qr [AID] [EID], and the simultaneous I / O presence rate Rr [AID] [EID].

[First Modification]
Next, a first modification of the above embodiment will be described with reference to FIG. The feature of the first modification is that the conditions for selecting the first array and the first extent that are subject to extent migration, and the second array and the second extent that are counterparts to the extent migration are as described above. There are some differences from the form.

  First, it is assumed that the arrays 11-1 (# 1), 11-2 (# 2), and 11-3 (# 3) are data rearrangement candidates (optimization candidates) by extent migration. A group consisting of these optimization candidates (corresponding to the RAID group described above) is called an optimization target group (or data relocation target group). The first extent includes simultaneous I / O among the representative HDDs 110 (# 1) to 110 (# 3) of the arrays 11-1 (# 1) to 11-3 (# 3) included in the optimization target group. The extent to which the extent chunk with the highest O existence rate belongs is used. Therefore, the array to which the first extent belongs is used as the first array.

  In the example of FIG. 11, the extent chunk having the highest simultaneous I / O existence rate among the representative HDDs 110 (# 1) to 110 (# 3) of the arrays 11-1 (# 1) to 11-3 (# 3). Is an extent chunk 111 in the representative HDD 110 (# 1). That is, the representative HDD including the extent chunk 111 having the highest simultaneous I / O existence rate among the representative HDDs 110 (# 1) to 110 (# 3) is the representative HDD 110 (# 1) of the array 11-1 (# 1). ). The first extent is the extent 114 in the array 11-1 (# 1) to which the extent chunk 111 belongs.

  The second array includes the representative HDD 110 having the lowest HDD I / O presence rate among the representative HDDs 110 (# 1) to 110 (# 3) of the arrays 11-1 (# 1) to 11-3 (# 3). (That is, the array having the lowest HDD utilization rate) is used. For the second extent, the extent to which the extent chunk having the lowest extent chunk I / O existence rate in the representative HDD 110 of the second array belongs is used among the extents in the second array described above.

  In the example of FIG. 11, the representative HDD 110 having the lowest HDD I / O existence rate is the representative HDD 110 (# 2) of the array 11-2 (# 2), and therefore the second array is the array 11-2 (# 2). It is. In the representative HDD 110 (# 2) of the array 11-2 (# 2), the extent chunk having the lowest extent chunk I / O existence ratio is the extent chunk 115, and the second extent is the array 11 to which the extent chunk 115 belongs. -1 (# 2).

  Extent migration 117 between the first extent 114 in the first array 11-1 (# 1) and the second extent 116 in the second array 11-2 (# 2) satisfying such conditions. By performing the above, it is possible to optimize the storage system more effectively than in the above embodiment.

[Second Modification]
Next, a second modification of the above embodiment will be described. A feature of the second modification is that, in order to improve the performance of the HDD 110, in-array extent migration is performed, unlike the above embodiment. Hereinafter, this in-array extent migration will be described with reference to FIG.
In FIG. 12, extent chunks 131 and 132 are included in the representative HDD 110 of the array 11-i. The extent chunks 131 and 132 store data D1 and D2, respectively. Here, it is assumed that the extent chunks 131 and 132 both have an executing I / O degree (extent chunk I / O existence rate) higher than a certain level. In this case, in the representative HDD 110, the degree of occurrence of a seek operation for moving the head from the extent chunk 131 to the extent chunk 132 or vice versa is high.

  However, since the extent chunks 131 and 132 are physically separated on the disk medium of the representative HDD 110, the distance that the head moves in the seek operation between the extent chunks 131 and 132 (seek distance) is long. It takes a long time (seek time). In such a state, the performance of the HDD 110 deteriorates due to a seek operation between the extent chunks 131 and 132 that occurs frequently.

  Therefore, the array controller 12-j (internal RAID manager 125) sets the extent to which one of the extent chunks 131 and 132 having a high extent chunk I / O existence ratio, for example, the extent to which the extent chunk 132 belongs, as the extent migration target in the array. select. The array controller 12-j (the RAID manager 125 in the array controller 12-j) is located in the vicinity of the other extent chunk 131 of the extent chunks 131 and 132, for example, adjacent to the extent chunk 131, and the extent chunk I / The extent to which the extent chunk 133 having a low O existence rate belongs is selected as the extent migration target in the array with the extent to which the extent chunk 132 belongs.

  Here, the positions of the extent chunks 131 and 133 (extents to which they belong) in the representative HDD 110 (array 11-i) between the extent chunks 131 and 132 (extents to which the extent chunks 131 and 132 (to which the extents belong) belong within the representative HDD 110 (array 11-i) of each other. Note that it is closer than the location.

  The array controller 12-j (within the RAID manager 125) selects between the selected pair of extents, that is, the extent to which the extent chunk 132 belongs and the extent to which the extent chunk 133 (adjacent to the extent chunk 131) belongs. In-array extent migration 134 is performed.

  As a result of the data rearrangement 135 by the in-array extent migration 134, the data D2 of the extent chunk 132 in the representative HDD 110 is rearranged in the extent chunk 133 in the representative HDD 110. In this extent migration 134, the data of another extent chunk included in the extent (third extent) to which the extent chunk 132 belongs is stored in another extent included in the extent (fourth extent) to which the extent chunk 133 belongs. Relocated to extent chunk.

  In the second modification, as a result of the data rearrangement 135 by the extent migration in the array 134, that is, as a result of the data D2 being rearranged from the extent chunk 132 to the extent chunk 133 adjacent to the extent chunk 131, the data D2 is stored. The extent chunk being located is closer to the extent chunk 131 in which the data D1 is stored on the HDD 110 than before the data relocation 135 is executed. As a result, the seek distance between the position of the extent chunk storing the data D1 and the position of the extent chunk storing the data D2 is shortened. That is, the seek time required to move the head from the extent chunk storing the data D1 to the extent chunk storing the data D2 or vice versa can be shortened. This is because the seek distance between the extent chunks 131 and 132 where the data D1 and D2 are stored before the data rearrangement 135 is shortened, and the seek time required for head movement between the extent chunks 131 and 132 is reduced. Is equivalent to shortening. The same applies to other extent chunks included in the extents to which the extent chunks 131 and 132 belong, respectively. As a result, the performance of each HDD 110 in the array 11-i is improved.

Next, conditions for executing in-array extent migration (by extent rearrangement) applied in the second modification will be described.
In this modification, the array controller 12-j (inside the RAID manager 125) has the following two conditions 1) The time elapsed since the last extent rearrangement is longer than a certain time 2) The extent migration candidates in the array The difference between the extent chunk I / O existence ratios of a pair of extent chunks included in the representative HDD of the array to which the pair of extents belongs is greater than a certain value. Extent relocation is executed when all of the above are satisfied.

  Here, a function f (Δ) that uniquely determines the execution cycle T according to the above-described difference Δ of the extent chunk I / O existence rate as the “certain time” in condition 1), that is, the extent relocation execution cycle. If the difference is larger than a certain level, the extent rearrangement cycle should be shortened. In this way, performance can be improved efficiently. As the function f (Δ), for example, when Δ is less than a certain level, f (Δ) indicates the first execution cycle T1, and when Δ is greater than or equal to a certain level, f (Δ) is shorter than the first execution cycle T1. What shows the execution cycle T2 may be used. It is also possible to use a function f (Δ) that is determined such that when Δ is greater than or equal to a certain level, the Δ becomes shorter as Δ increases.

  As described above, in the second modification in which intra-array extent migration is performed but inter-array extent migration is not performed, the simultaneous I / O existence rate is not necessary for the autonomous optimization control. In this case, the I / O statistical information collection module 125a in the RAID manager 125 only needs to calculate the HDD I / O existence rate Pr [AID] and the extent chunk I / O existence rate Qr [AID] [EID]. Therefore, the module 125a only needs to collect the HDD I / O existence sample number P [AID] and the extent chunk I / O existence sample number Q [AID] [EID].

  FIG. 13 shows a procedure in one sample (sample period) of the process of collecting the HDDI / O existence sample number P [AID] and the extent chunk I / O existence sample number Q [AID] [EID]. In FIG. 13, the same parts as those in FIG.

  In the second modified example, it is assumed that the number N of extent chunks whose extent chunk I / O existence ratio is higher than a certain level in the representative HDD 110 of the array 11-i is 2. However, the same applies to the case where N is 3 or more. That is, in-array extents between N-1 extents of N extents to which N extent chunks belong and N-1 extent chunks in the vicinity of the remaining one extent chunk, respectively. Migration should be performed. Further, as in the following third modification, all of the N extents to which N extent chunks having a high extent chunk I / O existence ratio belong can be targeted for extent migration in the array.

[Third Modification]
Next, a third modification of the above embodiment will be described with reference to FIG. The first feature of the third modified example is that the seek distances between the plurality of extent chunks are targeted for the extents to which a plurality of extent chunks having a high extent chunk I / O existence ratio belong in the representative HDD 110 of the array 11-i. The extent migration in the array is performed so that is equivalently shorter. The second feature of the third modification is that each extent (a plurality of extent chunks) to be subject to in-array extent migration is an extent chunk I / of the extent chunk in the representative HDD 110 included in the extent. In the descending order of the O existence rate, the HDDs 110 constituting the array 11-i are rearranged in order from the outer periphery of the disk 110 (a disk medium mounted on the HDD 11). In general, in the HDD 110, the logical block address (Logical Block Address: LBA) becomes smaller toward the outer peripheral side of the disk medium.

  In FIG. 14, the extent chunks 141 and 142 included in the representative HDD 110 of the array 11-i are assumed to have the first and second extent chunk I / O existence rates in the representative HDD 110, respectively.

  The array controller 12-j (internal RAID manager 125) first installs the extent to which the extent chunk I / O existence ratio of the first extent chunk I / O in the representative HDD 110 of the array 11-i belongs, and the representative HDD 110. In-array extent migration 145 is performed with the extent to which the extent chunk 143 located on the outermost periphery of the disk medium).

  Next, the array controller 12-j (inside the RAID manager 125) is adjacent to the extent to which the extent chunk I / O existence rate in the representative HDD 110 belongs and the extent chunk 143 in the representative HDD 110. In-array extent migration 146 is performed between the extent chunk 144 and the extent chunk 144 located on the inner peripheral side of the representative HDD 110 with respect to the extent chunk 143.

  By such extent migrations 145 and 146 in the array, the extents (a plurality of extent chunks) to which the extent chunks 141 and 142 having a high extent chunk I / O existence ratio in the representative HDD 110 belong are included in the extents. The extent chunk I / O existence ratios of extent chunks in the representative HDD 110 are rearranged in order from the outermost periphery of each HDD 110 (disk medium installed in the array 11-i) in descending order. For example, in the representative HDD 110, the data of the extent chunk 141 having the first extent chunk I / O existence rate is relocated to the outermost periphery (extent chunk 143) of the representative HDD 110, and the extent chunk I / O existence rate is the first. The data of the second extent chunk 142 is rearranged at a position adjacent to the rearrangement destination of the extent chunk 141 (extent chunk 144).

  In the third modification, the seek distance between the extent chunks 141 and 142 having the upper extent chunk I / O existence ratio is equivalently shortened by the extent migrations 145 and 146 in the array, and the extent chunks 141 and 142 are shortened. The seek time can be shortened. The same applies to other extent chunks included in the extents to which the extent chunks 141 and 142 belong, respectively. Moreover, in the third modified example, extent chunks having a higher extent chunk I / O existence rate are rearranged on the outer peripheral side of the disk medium mounted on the HDD 110. In general, the disk medium mounted on the HDD 110 has a larger number of sectors per track toward the outer peripheral side, and therefore the performance becomes higher. Therefore, in the third modification, the I / O performance of the HDD 110 can be further improved by the above-described extent migration in the array.

  Next, FIG. 15 shows a case where a plurality of extent chunks having the same I / O existence rate exist in the extent chunk group having the higher extent chunk I / O existence rate in the representative HDD 110 of the array 11-i. The description will be given with reference.

  In FIG. 15, extent chunks 151 and 152 have the same extent chunk I / O existence rate in the representative HDD 110, for example, both are ranked first. Further, it is assumed that the extent chunk 151 out of the extent chunks 151 and 152 is arranged on the outer peripheral side of the representative HDD 110. Furthermore, it is assumed that extents (internal extent chunks) constituting the array 11-i are physically arranged on the outer peripheral side of the HDD 110 as the extent number is smaller.

  The array controller 12-j (inside the RAID manager 125), when an extent to which each of a plurality of extent chunks having the same extent chunk I / O existence ratio is included in the in-array extent migration target, In-array extent migration is executed in the following order. That is, the array controller 12-j (internal RAID manager 125) preferentially performs in-array extent migration for extents having a larger extent number (extents farther from the outer periphery of the HDD 110).

  In the example of FIG. 15, first, the array controller 12-j (inside the RAID manager 125), the extent to which the extent chunk 152 of the extent chunks 151 and 152 belongs and the outer peripheral side (here, the outermost periphery) in the representative HDD 110. The in-array extent migration 155 is performed with the extent to which the extent chunk 153 located at the position belongs. Next, the array controller 12-j (inside the RAID manager 125) transfers the in-array extent migration 156 between the extent to which the extent chunk 151 belongs and the extent to which the extent chunk 154 adjacent to the extent chunk 153 in the representative HDD 110 belongs. I do. Here, the extent to which the extent chunk 152 belongs is arranged at a position (position closer to the inner periphery of the HDD 110) where the I / O performance is lower than the extent to which the extent chunk 151 belongs. Therefore, by giving priority to extent migration within the extent chunk 152, I / O performance can be further improved.

  Also in the third modified example, as in the second modified example, the number of HDD I / O existing samples P [AID] and the number of extent chunk I / O existing samples Q [AID] [EID] are set according to the flowchart of FIG. Processing to collect is performed.

  Next, in-array extent migration in the third modification will be described in more detail with reference to FIG. In FIG. 16, the array 11-i includes a plurality of HDDs 110 (for example, six HDDs 110 as in the above embodiment). The storage area of the array 11-i is managed by being divided into a plurality of extents (for example, 1500 extents as in the above embodiment). Here, EIDs are assigned to a plurality of extents in the order of 0, 1,... From the outer periphery side of the HDD 110.

In the third modification, the extent chunk I / O existence ratio Qr [AID] [EID] for each extent chunk in the representative HDD 110 in the array indicated by AID is sorted in descending order of the existence ratio. This sorting is performed by the autonomous optimization control engine 125c in the RAID manager 125, for example. The total number of extent chunks in the representative HDD 110 is represented by n (that is, MaxEID = n−1), and the Nth (N = 0, 1,..., N−1) Qr [AID] [EID] EID after sorting is represented by XN. And n Qr [AID] [EID], that is, Qr [AID] [X0], Qr [AID] [X1], ..., Qr [AID] [Xn-1]
Qr [AID] [X0] ≥Qr [AID] [X1] ≥ ... ≥Qr [AID] [Xn-1]
It becomes.

  Now, it is assumed that a pair of extents (extent migration execution pair) to be subjected to intra-array extent migration is a third extent 161 and a fourth extent 162 as shown in FIG. The third extent 161 and the fourth extent 162 are defined as follows.

  First, the fourth extent 162 is an extent to which the kth extent chunk from the outer periphery of the representative HDD 110 (a disk medium mounted on the representative HDD 110) belongs (that is, an extent with EID = k). The third extent 161 is an extent (extent of EID = Xk) to which an extent chunk of the kth largest extent chunk I / O existence rate Qr [AID] [Xk] in the representative HDD 110 belongs.

The condition (extent migration execution condition) under which in-array extent migration is executed between the third extent 161 and the fourth extent 162, as is clear from the conditions given in the second modification example,
1) Elapsed time since the last extent rearrangement is a certain time or more 2) Extent chunk I / O existence ratio Qr [AID] [Xk] of extent chunks in the representative HDD of the array to which the third and fourth extents belong And the difference between Qr [AID] [k] (Qr [AID] [Xk] −Qr [AID] [k]) is greater than or equal to a certain value.

  The autonomous optimization control engine 125c in the RAID manager 125 includes a pair of extents (physical extents) 161 and 162 that satisfy the above-described extent migration execution condition, that is, the third extent 161 of EID = Xk and the first extent of EID = k. 4, the extent migration between the extents 161 and 162 is instructed to the extent migration execution module 125b. Upon receiving this instruction, the module 125b performs in-array extent migration between the third extent 161 with EID = Xk and the fourth extent 162 with EID = k. This extent migration is performed sequentially by selecting the third extent in descending order of Qr [AID] [Xk].

[Fourth Modification]
Next, a fourth modification of the above embodiment will be described with reference to FIG. The feature of the fourth modification is a combination of the extent migration between arrays applied in the embodiment or the first modification and the extent migration in the array applied in the second or the third modification. It is in the point.

  For example, the array controller 12-1 (internal RAID manager 125), as the first step, extent migration 171 between the arrays 11-1 (# 1), 11-2 (# 2), and 11-3 (# 3) Perform extent relocation by. By this extent rearrangement, the array controller 12-1 (inside the RAID manager 125) allows each of the representative HDDs 110 in the arrays 11-1 (# 1), 11-2 (# 2), and 11-3 (# 3). The HDD I / O presence rate is equalized to a certain level.

  Next, as a second step, the array controller 12-1 (inside the RAID manager 125) performs the following operations on the arrays 11-1 (# 1), 11-2 (# 2), and 11-3 (# 3). Extent rearrangement is performed by in-array extent migrations 172-1, 172-2, and 172-3, respectively. By this extent rearrangement, the array controller 12-1 (internal RAID manager 125) allows the HDDs 110 constituting the arrays 11-1 (# 1), 11-2 (# 2), and 11-3 (# 3) to be Reduce seek time.

[Fifth Modification]
Next, a fifth modification of the above embodiment will be described with reference to FIGS. The feature of the fifth modification is that the extent migration between arrays applied in the embodiment or the first modification is applied between different disk array apparatuses.

  As a first step, the array controller 12-1 (inside the RAID manager 125), as shown in FIG. 18A, the arrays 11-1 (# 1), 11-2 (in the disk array apparatus 10-1) Extent rearrangement is performed by extent migration 181-1 between # 2) and 11-3 (# 3). By this extent rearrangement, the array controller 12-1 (inside the RAID manager 125) allows each of the representative HDDs 110 in the arrays 11-1 (# 1), 11-2 (# 2), and 11-3 (# 3). The HDD I / O presence rate is equalized to a certain level.

  On the other hand, the array controller 12-2 (inside the RAID manager 125) also has, as a first step, as shown in FIG. 18A, the arrays 11-A (#A), 11- in the disk array device 10-2. Extent rearrangement is performed by extent migration 181-2 between B (#B) and 11-C (#C). By this extent rearrangement, the array controller 12-2 (inside the RAID manager 125) allows each of the representative HDDs 110 in the arrays 11-A (#A), 11-B (#B), and 11-C (#C). The HDD I / O presence rate is equalized to a certain level.

  As described above, in the first step, the extent migrations 181-1 and 181-2 between the arrays are performed in the disk array apparatuses 10-1 and 10-2, respectively, and the disk array apparatuses 10-1 and 10-2 Optimization is performed.

  Next, the array controllers 12-1 and 12-2 (inside the RAID manager 125) communicate with each other as the second step, and as shown in FIG. 18B, the arrays in the disk array device 10-1 11-1 (# 1), 11-2 (# 2) and 11-3 (# 3), and arrays 11-A (#A), 11-B (#B) in the disk array device 10-2 and 11-C (#C), that is, the extent rearrangement by the extent migration 182 between the arrays via the network 32 between the disk array apparatuses 10-1 and 10-2. By this extent rearrangement, the array controllers 12-1 and 12-2 (inside the RAID manager 125) allow the arrays 11-1 (# 1), 11-2 (# 2), and 11 in the disk array device 10-1. -3 (# 3) and the presence of HDD I / O in the representative HDD 110 of each of the arrays 11-A (#A), 11-B (#B), and 11-C (#C) in the disk array device 10-2 Equalize the rate to a certain level.

  As described above, in the second step, the extent migration 182 between the arrays is performed between the disk array devices 10-1 and 10-2, and the optimization between the disk array devices 10-1 and 10-2 is performed. Is called.

  After this second step, the array controller 12-1 (internal RAID manager 125), as the third step, as shown in FIG. 19, arrays 11-1 (# 1), 11-2 (# 2 ) And 11-3 (# 3), extent rearrangement is performed by in-array extent migration 191-1, 191-2, and 191-3, respectively.

  On the other hand, as shown in FIG. 19, the array controller 12-2 (internal RAID manager 125) also has arrays 11-A (#A), 11-B (#B), and 11-C (# For C), extent rearrangement is performed by in-array extent migration 191-A, 191-B, and 191-C, respectively. Note that the third step may be omitted.

  In addition, this invention is not limited to the said embodiment or its modification example as it is, A component can be deform | transformed and embodied in the range which does not deviate from the summary in an implementation stage. For example, in-execution I / O existence rates in array units and extent units may be used instead of in-execution I / O existence rates in HDD units and extent chunk units. That is, as the first existence rate, the degree of simultaneous execution of I / O in a plurality of extents in the array is used, and as the second existence ratio, the degree of execution of I / O in the array is used. It is also possible to use, as the third existence ratio, the degree of presence of executing I / O in the extent. In addition, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment or the modification thereof. For example, you may delete a some component from all the components shown by embodiment or its modification.

1 is a block diagram showing a configuration of a disk array subsystem according to an embodiment of the present invention. The figure which shows the relationship between the extent and extent chunk in an array. The block diagram which shows the structure of the array controller shown by FIG. The figure for demonstrating the extent migration in the same embodiment. The figure for demonstrating the reason for using simultaneous I / O existence rate to determine the extent used as extent migration object. The figure for demonstrating the method of acquiring the information used as the parameter | index showing a simultaneous I / O presence rate in the embodiment. The figure for demonstrating the method of acquiring the information used as the parameter | index showing HDDI / O presence rate in the embodiment. The figure for demonstrating the method of acquiring the information used as the parameter | index showing the extent chunk I / O presence rate in the embodiment. FIG. 3 is a block diagram showing a main configuration of a RAID manager in the array controller shown in FIG. 2 in association with other elements in the array controller. 6 is a flowchart showing a procedure in one sample of processing for collecting the number of HDD I / O existing samples, the number of extent chunk I / O existing samples, and the number of simultaneous I / O existing samples applied in the embodiment. The figure for demonstrating the extent migration in the 1st modification of the said embodiment. The figure for demonstrating the extent migration in the array in the 2nd modification of the said embodiment. The flowchart which shows the procedure in 1 sample of the process which collects the HDDI / O presence sample number and extent chunk I / O presence sample number applied in the said 2nd modification. The figure for demonstrating the extent migration in the array in the 3rd modification of the said embodiment. The figure for demonstrating the extent migration in an array in the case of the said 3rd modification in case there exist several extent chunks with equal I / O presence rate. The figure for demonstrating the detail of the extent migration in an array in the 3rd modification. The figure for demonstrating the combination of the extent migration between arrays applied in the 4th modification of the said embodiment, and the extent migration in an array. The figure for demonstrating the combination of the extent migration between arrays performed for every disk array apparatus applied in the 5th modification of the said embodiment, and the extent migration between arrays performed between disk array apparatuses. The figure for demonstrating the extent migration in the array applied in addition to the combination shown in FIG. 18 in the said 5th modification.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Disk array subsystem, 10-1, 10-2 ... Disk array apparatus, 11-1 to 11-3, 11-A to 11-C ... Array, 12-1, 12-2, 12-j ... Array Controller 20 ... Host device 31, 32 ... Network 110 ... HDD 125 ... RAID manager 125a ... I / O statistical information collection module 125b ... Extent migration execution module 125c ... Autonomous optimization control engine 126 ... HDD Driver, 127 ... memory.

Claims (10)

The storage area of one or more disk devices includes a plurality of arrays defined as one continuous area, the entire area of the array is divided into a plurality of extents, and one or more extents are combined to form a host device. In a disk array subsystem having at least one disk array device constituting a recognized logical volume,
The at least one disk array device includes:
For each of the arrays, an incomplete access request is executed at a sampling cycle for each extent chunk belonging to each of the plurality of extents included in a storage area of a predetermined representative disk device among the disk devices constituting the array. By sampling as a medium I / O, a first presence rate indicating the degree of simultaneous execution of I / O in a plurality of extent chunks including the extent chunk, and the execution of the I / O in the representative disk device Calculating means for calculating a second existence ratio indicating a degree of performing, and a third existence ratio indicating a degree of executing I / O in the extent chunk;
Based on the first to third existence ratios, an extent to which an extent chunk whose first existence ratio is to be reduced belongs is selected as a first extent, and a representative disk device of an array including the first extent The third existence from among extent chunks included in the representative disk device of the array having a relatively low second existence rate with respect to another representative disk device of the array including the first extent. A selection means for selecting the extent to which the extent chunk having the lowest rate belongs, as a second extent;
A disk array subsystem, comprising: extent migration executing means for executing inter-array extent migration for rearranging data between the first and second extents. The selection means includes
Among the plurality of arrays, the extent to which the extent chunk having the highest first existence rate belongs to the first chunk among the extents that constitute each of the arrays belonging to the array group predetermined as a candidate for data rearrangement. Select as extent,
The extent chunk with the lowest third existence rate is selected from the extent chunks included in the representative disk device with the lowest second existence ratio among the representative disk devices of each array belonging to the array group. The disk array subsystem according to claim 1, wherein the extent to which the selected extent chunk belongs is selected as the second extent.   The selection means has an elapsed time since the extent migration was last executed by the extent migration execution means is a predetermined time or more, and the maximum value of the second existence ratio of the representative disk devices of the array belonging to the array group And the minimum value of the second existing ratio of the array representative disk devices belonging to the array group is equal to or less than a predetermined value, and the difference between the minimum value and the minimum value is included in the representative disk device of the array belonging to the array group. 3. The first and second extents to be subjected to the next extent migration are selected when the maximum value of the first existence rate of extent chunks to be obtained becomes a certain value or more. The described disk array subsystem. If the selection means assumes that N is an integer of 2 or more, the third means can select any one of the plurality of arrays from among the extent chunks included in the representative disk device of the arbitrary array. N extent chunks having a high presence rate are selected, and N extents to which the selected N extent chunks belong or N-1 extents of the N extents are counted one by one. 3 extents, and N or N-1 extents in the arbitrary array to be subjected to in-array extent migration with the N or N-1 third extents. In the array, the positions of the N extents that are subject to extent migration in the array, One of the N extents excluding the N-1 third extent and the position in the array of the N-1 extents that are subject to the extent migration in the array are in the array. N or N−1 extents closer to the positions in the array of the N extents to which the N extent chunks having a high presence rate of 3 belong are selected as fourth extents one by one. ,
The extent migration execution means executes the extent migration in the array in which data is rearranged between the selected third and fourth extents every time the pair of the third and fourth extents is selected. 2. The disk array subsystem according to claim 1, wherein: The selecting means selects extents constituting an arbitrary array of the plurality of arrays one by one in descending order of the third existence ratio from the extent to which the extent chunk having the highest third existence ratio belongs. The extents that are subject to in-array extent migration between the selected third extents are selected one by one from the outer periphery to the inner periphery of the disk device constituting the arbitrary array. By sequentially selecting the third and fourth extent pairs to be subject to the in-array extent migration,
The extent migration execution means executes in-array extent migration that rearranges data between the selected third and fourth extents each time the pair of the third and fourth extents is selected. The disk array subsystem according to claim 1.   The selection means includes an elapsed time after the last extent migration in the array is executed by the extent migration execution means for a predetermined time or more, and is included in a pair of extents that are candidates for the extent migration in the array. When the difference between the third existence ratios of a pair of extent chunks included in the representative disk device of the array to which the pair of extents belongs is equal to or greater than a certain value, the pair of extent chunks 6. The disk array subsystem according to claim 4, wherein the extent is selected as a pair of the third and fourth extents to be subject to the next in-array extent migration.   The extent migration execution means executes extent migration between the arrays, and then executes extent migration in the array for each of the arrays that are targets of extent migration between the arrays. The disk array subsystem according to claim 4 or 5. The at least one disk array device is a plurality of disk array devices connected by a network;
The extent migration execution means of the plurality of disk array devices performs extent migration between the plurality of disk array devices after performing extent migration between the arrays in the disk array device, 6. The disk array subsystem according to claim 4, wherein extent migration in the array is executed for each of the arrays that are subject to extent migration between the arrays thereafter. The storage area of one or more disk devices includes an array defined as one continuous area, and the entire area of the array is divided into a plurality of extents and is recognized by the host device by combining one or more extents. In a disk array subsystem having at least one disk array device constituting a logical volume,
The at least one disk array device includes:
Of the extent chunks belonging to the plurality of extents included in the storage area of a predetermined representative disk device among the disk devices constituting the array, an uncompleted access request is executed as an executing I / O in a certain sampling period. A calculating means for calculating an extent chunk I / O existence ratio representing a degree of executing I / O in the extent chunk by sampling;
Assuming that N is an integer of 2 or more, the third existence rate is high among extent chunks included in the representative disk device of the arbitrary array for an arbitrary array of the plurality of arrays. Select N extent chunks, and select N extents to which the selected N extent chunks belong or N-1 extents of the N extents as third extents one by one. And N or N-1 extents in the arbitrary array subject to the extent migration in the array between the N or N-1 third extents, and in the array The position in the array of the N extents subject to extent migration, or the N extents. One of the tents excluding the N-1 third extent and the position in the array of the N-1 extents to be subject to the extent migration in the array is the third existence rate. Selecting means for selecting N or N-1 extents, one by one, closer to the positions in the array of the N extents to which the N extent chunks having higher
Extent migration executing means for executing in-array extent migration for rearranging data between the selected third and fourth extents each time the third and fourth extent pairs are selected. A disk array subsystem characterized by The storage area of one or more disk devices includes a plurality of arrays defined as one continuous area, the entire area of the array is divided into a plurality of extents, and one or more extents are combined to form a host device. A program executed by a controller included in a disk array device constituting a recognized logical volume,
In the controller,
For each of the arrays, an incomplete access request is executed at a sampling cycle for each extent chunk belonging to each of the plurality of extents included in a storage area of a predetermined representative disk device among the disk devices constituting the array. By sampling as a medium I / O, a first presence rate indicating the degree of simultaneous execution of I / O in a plurality of extent chunks including the extent chunk, and the execution of the I / O in the representative disk device Calculating a second existence ratio representing a degree of performing and a third existence ratio representing a degree of executing I / O in the extent chunk;
Based on the first to third existence ratios, an extent to which an extent chunk whose first existence ratio is to be reduced belongs is selected as a first extent, and a representative disk device of an array including the first extent The third existence from among extent chunks included in the representative disk device of the array having a relatively low second existence rate with respect to another representative disk device of the array including the first extent. Selecting the extent to which the extent chunk with the lowest rate belongs as a second extent;
A program for executing an inter-array extent migration that rearranges data between the first and second extents.

Images