JP7518364B2 - Information processing device and path control method - Google Patents

Description

The present invention relates to an information processing device and a path control method.

By making the access paths that computers use to access storage devices redundant, it is possible to increase availability in the event of a failure and distribute the communication load. With regard to availability, one example of access path redundancy is to provide an active path that is used during normal operation, and a standby path that is used when a failure occurs in the active path.

The following proposals have been made regarding such multipath control. For example, a storage server has been proposed in which multipath programming is performed using a port failover policy that is used to intelligently adapt to compatible storage devices, including active-active and active-passive policies. Also, a computer system has been proposed that, when an error occurs with an issued I/O (Input/Output), starts supplying power to a standby HBA (Host Bus Adaptor) and then reissues the same I/O to the active path, and if no error occurs in response to this, stops supplying power to the standby HBA after a specified time.

International Publication No. WO 2002/065290 JP 2010-198353 A

In some cases, the active path and the standby path to the storage device each include multiple paths. In this configuration, normally, if a failure occurs in one of the paths included in the active path, the remaining paths included in the active path are used to access the storage device. However, in this case, the communication load on each of the remaining paths increases, causing a problem of a significant decrease in access performance.

In one aspect, the present invention aims to provide an information processing device and a path control method that can suppress a decrease in access performance when a failure occurs in one of the paths included in the active path.

In one proposal, an information processing device is provided that has a communication unit and a processing unit as follows. In this information processing device, the communication unit connects to a storage device via a plurality of first paths of an active system and a plurality of second paths of a standby system. When some of the plurality of first paths become unavailable, the processing unit determines whether to distribute access requests to the storage device to the remaining first paths or to the remaining first paths and the plurality of second paths based on the communication load for each of the remaining available first paths and the plurality of second paths among the plurality of first paths and the processing capacity assigned by the storage device.

In one proposal, a path control method is provided in which a computer executes processing similar to that of the information processing device described above.

In one aspect, it is possible to suppress a decrease in access performance when a failure occurs in one of the paths included in the active path.

1 illustrates a configuration example and a processing example of a storage system according to a first embodiment; FIG. 1 illustrates a first configuration example of a storage system. FIG. 2 illustrates an example of a hardware configuration of a server. FIG. 2 is a diagram illustrating an access path in a first configuration example. FIG. 11 is a diagram illustrating a second configuration example of a storage system. FIG. 13 is a diagram illustrating an access path in a second configuration example. FIG. 2 is a diagram illustrating an example of a configuration of processing functions provided in a server and a CM. FIG. 2 is a diagram for explaining a resource allocation function of a storage device. 11 is an example of a flowchart illustrating a resource allocation process according to a server connection. 13 is a first example of a flowchart illustrating a resource allocation process in response to an I/O request. 13 is a second example of a flowchart illustrating a resource allocation process in response to an I/O request. FIG. 13 is a diagram illustrating an example of the configuration of a performance management table of a CM. FIG. 2 illustrates an example of the configuration of a performance management table of a server. FIG. 13 is a diagram illustrating an example of a normal state. FIG. 11 illustrates a first example of a state in which a failure occurs in an active path. FIG. 11 illustrates a second example of a state in which a failure occurs in an active path. FIG. 13 illustrates a third example of a state in which a failure occurs in an active path. FIG. 13 is a diagram illustrating an example of a normal state. FIG. 13 illustrates a fourth example of a state in which a failure occurs in an active path. FIG. 13 illustrates a fifth example of a state in which a failure occurs in an active path. 13 is a diagram showing estimated values of performance information when an active path and a standby path are used in combination. FIG. 13 is an example of a flowchart illustrating a path selection process when a failure occurs. 13 is an example of a flowchart illustrating a path selection process when performance information is acquired.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
First Embodiment
1 is a diagram showing an example of the configuration and processing of a storage system according to a first embodiment of the present invention. The storage system shown in FIG.

The information processing device 10 includes a communication unit 11 and a processing unit 12. The communication unit 11 is a communication interface for communicating with the storage device 20. The processing unit 12 is realized, for example, as a processor. On the other hand, the storage device 20 includes a memory area 21 realized by a storage device such as a hard disk drive (HDD) or a solid state drive (SSD). The processing unit 12 of the information processing device 10 accesses the memory area 21 by sending an access request to the storage device 20 via the communication unit 11.

In the example of FIG. 1, paths 1a, 1b, 2a, and 2b are provided as access paths for the processing unit 12 to access the storage area 21. In FIG. 1, as an example, the communication unit 11 has ports P1 to P4, and the storage device 20 has ports P11 to P14. Path 1a connects port P1 to port P11, path 1b connects port P2 to port P12, path 2a connects port P3 to port P13, and path 2b connects port P4 to port P14.

Of these paths, paths 1a and 1b are used as active paths, and paths 2a and 2b are used as standby paths. The processing unit 12 has a function of controlling through which of paths 1a, 1b, 2a, and 2b an access request is sent. When both active paths 1a and 1b are in a state where communication is possible normally, the processing unit 12 controls so that access requests are distributed to paths 1a and 1b. In this state, the standby paths 2a and 2b are not used. Furthermore, when both paths 1a and 1b become unavailable due to a failure, the processing unit 12 controls so that access requests are distributed to standby paths 2a and 2b.

On the other hand, if access requests are distributed to only the remaining usable paths when only one of the active paths 1a and 1b becomes unavailable, the access requests will be concentrated on the remaining paths, resulting in a significant degradation of access performance. For example, if access requests could be sent using the standby paths 2a and 2b in addition to the remaining active paths, the communication load could be distributed and degradation of access performance could be suppressed.

However, the communication performance of the standby paths 2a and 2b is not necessarily the same as that of the active path. For example, the storage device 20 may be equipped with multiple control devices that control access to the storage area 21. In this configuration, for example, the path using the active paths 1a and 1b reaches the storage area 21 via only one of the multiple control devices (the first control device). On the other hand, the path using the standby paths 2a and 2b may reach the storage area 21 via another of the multiple control devices (the second control device) and then via the first control device. In this case, the response speed when an access request is sent to the standby paths 2a and 2b is slower than when the active paths 1a and 1b are used.

Alternatively, for example, ports P13 and P14 to which the standby paths 2a and 2b are connected may also be used as access paths (e.g., active paths) from other information processing devices that access the storage device 20. Even in this case, depending on the communication conditions from the other information processing devices, the response speed when an access request is sent to the standby paths 2a and 2b may be slower than when the active paths 1a and 1b are used.

When one of the active paths 1a, 1b becomes unavailable, the processing unit 12 selects an appropriate path and distributes access requests based on the communication load 3a and processing capacity 3b for the remaining path and the standby paths 2a, 2b. Here, the processing capacity 3b is assigned to each of the above paths by the storage device 20. For example, the processing capacity is assigned to each path based on the usage status of the ports on the storage device 20 side of each of the above paths as active paths and standby paths, the communication load at those ports, etc. The processing unit 12 acquires the processing capacity 3b assigned to each of the paths 1a, 1b, 2a, 2b, for example, periodically from the storage device 20. In addition, the communication load 3a may be acquired as a detected value from the storage device 20, or may be detected on the information processing device 10 side.

As shown in FIG. 1, it is assumed below that path 1a has become unavailable due to a failure. At this time, the processing unit 12 refers to the communication load 3a and processing capacity 3b for the remaining active path 1b that is available and the standby paths 2a and 2b. Then, based on the communication load 3a and processing capacity 3b, the processing unit 12 determines whether to distribute the access request to the remaining active path 1a (Pattern A) or to distribute the access request to path 1a and the standby paths 2a and 2b (Pattern B). This makes it possible to appropriately suppress the deterioration of access performance.

For example, the processing unit 12 can estimate, from the communication load 3a and the processing capacity 3b, which of pattern A and B should be selected to shorten the response time to an access request. Therefore, it is possible to appropriately suppress a decrease in access performance while efficiently using existing paths. Alternatively, the processing unit 12 may estimate which of pattern A and B should be selected to shorten the response time to an access request when it is estimated that distributing access requests to the remaining path 1a of the active system will significantly decrease the response time (for example, exceed a predetermined threshold value). In this case, it is possible to maintain a certain level of access performance while efficiently using existing paths.

Second Embodiment
Next, a storage system according to a second embodiment will be described. In the second embodiment, two system configurations, a "first configuration example" and a "second configuration example", will be illustrated.

Figure 2 is a diagram showing a first configuration example of a storage system. The storage system shown in Figure 2 includes a server 100 and a storage device 200. The storage device 200 includes CMs (Controller Modules) 201 and 202 and a drive unit 203. The server 100 and the CMs 201 and 202 are connected via a network (not shown). This network is, for example, a SAN (Storage Area Network) using FC (Fibre Channel).

The drive unit 203 is equipped with multiple non-volatile storage devices such as HDDs and SSDs. The CMs 201 and 202 are storage control devices that access the storage devices of the drive unit 203 in response to I/O requests from the server 100. The CMs 201 and 202 are capable of communicating with each other. In the following description, the CMs 201 and 202 may be referred to as CM#0 and CM#1, respectively. The server 100 is a server computer that executes predetermined processes such as business processes, and is able to access the drive unit 203 by sending I/O requests to the CMs 201 and 202.

Here, server 100 is equipped with HBA#0-#7. HBA#0-#7 are interfaces for communicating with storage device 200, and each has its own individual communication port. Meanwhile, CM 201 is equipped with ports #0-#3 as communication ports. Also, CM 202 is equipped with ports #0-#3 as communication ports. HBA#0-#3 of server 100 are connected to ports #0-#3 of CM 201, respectively, and HBA#4-#7 of server 100 are connected to ports #0-#3 of CM 202, respectively.

Figure 3 is a diagram showing an example of the hardware configuration of a server. The server 100 can be realized, for example, as a computer having the configuration shown in Figure 3. As shown in Figure 3, the server 100 includes a processor 101, a RAM (Random Access Memory) 102, a HDD 103, a graphic interface (I/F) 104, an input interface (I/F) 105, a reader 106, and a communication interface (I/F) 107.

The processor 101 performs overall control of the entire server 100. The processor 101 is, for example, a CPU (Central Processing Unit), an MPU (Micro Processing Unit), a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), or a PLD (Programmable Logic Device). The processor 101 may also be a combination of two or more elements of a CPU, an MPU, a DSP, an ASIC, or a PLD.

RAM 102 is used as the main storage device of server 100. RAM 102 temporarily stores at least a portion of the OS (Operating System) program and application programs to be executed by processor 101. RAM 102 also stores various data necessary for processing by processor 101.

The HDD 103 is used as an auxiliary storage device for the server 100. The HDD 103 stores the OS program, application programs, and various data. Note that other types of non-volatile storage devices, such as SSDs, can also be used as the auxiliary storage device.

A display device 104a is connected to the graphic interface 104. The graphic interface 104 displays an image on the display device 104a in accordance with an instruction from the processor 101. Examples of the display device include a liquid crystal display and an organic EL (ElectroLuminescence) display.

An input device 105a is connected to the input interface 105. The input interface 105 transmits a signal output from the input device 105a to the processor 101. Examples of the input device 105a include a keyboard and a pointing device. Examples of the pointing device include a mouse, a touch panel, a tablet, a touch pad, and a trackball.

A portable recording medium 106a is detachably attached to the reading device 106. The reading device 106 reads data recorded on the portable recording medium 106a and transmits it to the processor 101. Examples of the portable recording medium 106a include an optical disk, a magneto-optical disk, and a semiconductor memory.

The communication interface 107 transmits and receives data to and from other devices such as CMs 201 and 202 via the network 300. In the first configuration example shown in FIG. 2, HBAs #0 to #7 are installed as the communication interfaces 107.

The processing functions of server 100 can be realized with the hardware configuration described above. Note that CMs 201 and 202 can also be realized as computers with the configuration shown in FIG. 3.

Figure 4 is a diagram showing an access path in the first configuration example. First, the data area 211 shown in Figure 4 is a storage area realized by a storage device mounted in the drive unit 203, and is, for example, a logical volume managed by RAID (Redundant Array of Inexpensive Disks). In the storage device 200, for each data area realized by the drive unit 203, a "responsible CM" is determined that directly accesses the data area and executes I/O processing. In the first configuration example, CM#0 is assumed to be the responsible CM for the data area 211. In this case, direct I/O processing for the data area 211 is executed by CM#0.

The access path for accessing the data area 211 from the server 100 is a multipath configuration having an active path Pa and a standby path Ps. Basically, the active path Pa is the access path used during normal operation, and the standby path Ps is the access path used when an abnormality occurs in the active path Pa.

In the first configuration example, as shown in FIG. 4, the active path Pa is a path that travels from HBAs #0-#3 of server 100 to data area 211 via ports #0-#3 of CM#0. The standby path Ps is a path that travels from HBAs #4-#7 of server 100 to data area 211 via ports #0-#3 of CM#1.

However, as described above, the CM responsible for data area 211 is CM#0, so standby path Ps travels from CM#1 through CM#0 to reach data area 211. Therefore, when data area 211 is accessed using standby path Ps, communication between CM#0 and CM#1 (inter-CM communication) occurs, and the access speed is slower than when active path Pa is used.

In the example of FIG. 4, the active path Pa between the server 100 and CM#0 has four communication paths using four sets of communication ports, which provides redundancy to the communication paths. Similarly, the standby path Ps between the server 100 and CM#1 also has four communication paths using four sets of communication ports, which provides redundancy to the communication paths.

Next, a second configuration example will be described.
5 is a diagram showing a second configuration example of a storage system. The second configuration example differs from the first configuration example in that two servers 100a and 100b are connected to the storage device 200. The servers 100a and 100b are connected to the CMs 201 and 202 of the storage device 200 via a switch 310. Both of the servers 100a and 100b can access the drive unit 203 via the switch 310 and the CMs 201 and 202. In the following description, the servers 100a and 100b may be referred to as servers #0 and #1, respectively.

As shown in FIG. 5, server 100a has HBAs #0 to #3. Server 100b has HBAs #0 to #3. Switch 310 has ports #0 to #11. CM 201 has ports #0 and #1. CM 202 has ports #0 and #1.

HBAs #0 to #3 of server 100a are connected to ports #0 to #3 of switch 310, respectively, and HBAs #0 to #3 of server 100b are connected to ports #4 to #7 of switch 310, respectively. Ports #8 and #9 of switch 310 are connected to ports #0 and #1 of CM 201, and ports #10 and #11 of switch 310 are connected to ports #0 and #1 of CM 202.

Furthermore, in switch 310, ports #0 and #1 are connected to ports #8 and #9, respectively, and ports #2 and #3 are connected to ports #10 and #11, respectively. Also, ports #4 and #5 are connected to ports #8 and #9, respectively, and ports #6 and #7 are connected to ports #10 and #11, respectively.

Figure 6 shows the access paths in the second configuration example. Figure 6 (A) shows the access path from server #0 to the data area, and Figure 6 (B) shows the access path from server #1 to the data area.

Data area 212 shown in FIG. 6(A) and data area 213 shown in FIG. 6(B) are both storage areas realized by a storage device mounted in drive unit 203, and are, for example, logical volumes managed by RAID. In the second configuration example, data area 212 is the storage area to be accessed by server #0, and data area 213 is the storage area to be accessed by server #1. Also, in the second configuration example, CM #0 is the CM in charge of data area 212, and CM #1 is the CM in charge of data area 213. In this case, direct I/O processing for data area 212 is executed by CM #0, and direct I/O processing for data area 213 is executed by CM #1.

As shown in FIG. 6(A), the access path for accessing the data area 212 from server #0 is a multipath configuration having an active path Pa1 and a standby path Ps1. Basically, the active path Pa1 is the access path used during normal operation, and the standby path Ps1 is the access path used when an abnormality occurs in the active path Pa1.

Active path Pa1 is a path that travels from HBA #0 and #1 of server #0 to data area 212 via ports #0 and #1 of switch 310 (not shown), ports #8 and #9 of switch 310 (not shown), and ports #0 and #1 of CM #0. Standby path Ps1 is a path that travels from HBA #2 and #3 of server #0 to data area 212 via ports #2 and #3 of switch 310 (not shown), ports #10 and #11 of switch 310 (not shown), and ports #0 and #1 of CM #1.

However, as described above, the CM responsible for the data area 212 is CM#0, so the standby path Ps1 travels from CM#1 through CM#0 to reach the data area 212. Therefore, when the data area 212 is accessed using the standby path Ps1, inter-CM communication occurs.

On the other hand, as shown in FIG. 6(B), the access path for accessing the data area 213 from server #1 is a multipath configuration having an active path Pa2 and a standby path Ps2. Basically, the active path Pa2 is the access path used during normal operation, and the standby path Ps2 is the access path used when an abnormality occurs in the active path Pa2.

Active path Pa2 is a path that travels from HBA #2 and #3 of server #1 to data area 213 via ports #6 and #7 (not shown) of switch 310, ports #10 and #11 (not shown) of switch 310, and ports #0 and #1 of CM #1. Standby path Ps2 is a path that travels from HBA #0 and #1 of server #1 to data area 213 via ports #4 and #5 (not shown) of switch 310, ports #8 and #9 (not shown) of switch 310, and ports #0 and #1 of CM #0.

However, as described above, the CM responsible for data area 213 is CM#1, so standby path Ps2 travels from CM#0 through CM#1 to reach data area 213. Therefore, when data area 213 is accessed using standby path Ps2, inter-CM communication occurs.

In both the first and second configuration examples above, the active path from a server to a data area includes multiple physically independent transmission paths. For example, in the first configuration example, the active path Pa from the server 100 to the data area 211 includes transmission paths that pass through each of HBAs #0 to #3 of the server 100. In this way, when the transmission paths included in the active path are made redundant, even if a failure occurs in one of the transmission paths, I/O processing between the data area can be continued using the remaining surviving transmission paths. For this reason, I/O processing using the standby path is generally performed only after all transmission paths in the active path become unavailable.

However, if some of the transmission paths included in the active path become unavailable, all I/O processing is performed using only the remaining transmission paths, which causes an excessive communication load on each transmission path and a decrease in I/O processing performance. In addition, the standby path is not used until all transmission paths in the active path become unavailable, resulting in a waste of resources.

Therefore, in this embodiment, when at least one transmission path included in the active path becomes unavailable, the server is able to execute I/O processing using not only the remaining surviving transmission paths but also the standby path transmission paths. This makes effective use of the transmission path resources and minimizes degradation of I/O processing performance.

However, as mentioned above, I/O processing using a standby path involves communication between CMs, so when an individual transmission path included in the standby path is used, the response speed is slower than when an individual transmission path included in the active path is used. Also, as in the second configuration example, the communication port on the storage device 200 side of the standby path may also be used as an active path for another server. For these reasons, the extent to which I/O processing performance can be improved by using the standby path when some transmission paths included in the active path become unavailable varies depending on the situation.

Because a standby path is essentially a spare access path, it is not desirable to use it carelessly while some of the transmission paths of the active path are still alive. Furthermore, if part of the route of the standby path of a certain server overlaps with the active path of another server, as in the second configuration example, using the former path will affect the I/O processing performance of the latter path. Therefore, when some of the transmission paths of the active path are still alive, it is desirable to be able to control the standby path so that it is used only when there is a high need.

Here, the storage device 200 of this embodiment has a function (resource allocation control function) for controlling the allocation of resources (distribution ratio of processing amount) to I/O requests from each HBA of the server for each communication port of the CM according to the usage status by I/O processing. In this function, the processing capacity of I/O processing using the access path via the communication port is calculated based on the resource allocation amount and the detected value of the I/O processing load. Therefore, in this embodiment, such a function is used so that the storage device 200 can notify the server of the current I/O processing load and I/O processing capacity for each communication port of the CM as performance information. Then, when some transmission paths of the active path become unavailable, the server judges whether it is better to execute I/O processing using not only the remaining transmission paths of the active path but also the standby path based on the notified performance information. This makes it possible to suppress the deterioration of I/O processing performance due to the occurrence of a failure while balancing the resource usage.

Figure 7 is a diagram showing an example of the configuration of the processing functions of the server and CM. As an example, Figure 7 shows the processing functions of the server 100 and CMs 201 and 202 in the first configuration example.

First, CM201 includes a memory unit 220 and a RAID control unit 230. The memory unit 220 is realized as a storage area of a storage device included in CM201, such as a RAM or HDD. The RAID control unit 230 is realized, for example, by a processor included in CM201 executing a predetermined program.

The memory unit 220 stores a performance management table 221. In the performance management table 221, performance information is registered for each combination of a communication port provided in the CM 201 and an HBA of a server connected to the communication port. When multiple servers are connected to the CM 201 as in the second configuration example, a record is registered in the performance management table 221 for each combination of a communication port on the CM 201 side and each HBA of each connected server. Although not shown, the memory unit 220 also stores information indicating the correspondence between the server from which the access originates, the data area of the access destination, and the communication ports corresponding to the active path and the standby path.

The RAID control unit 230 executes I/O processing for the data area that is the target of I/O processing in response to an I/O request received from the server. At this time, the I/O processing is executed while managing the storage device of the drive unit 203 by RAID. Furthermore, when the RAID control unit 230 receives an I/O request via a standby path for a data area that is not managed by the CM 201 itself, it transfers the I/O request to the CM in charge.

The RAID control unit 230 further includes a resource allocation unit 231. The resource allocation unit 231 controls the allocation of resources to received I/O requests according to the resource usage status by the I/O processing. The "resource allocation" referred to here refers to the allocation of resources used in I/O processing according to the I/O request, and indicates, for example, the number of allocated processor cores and the amount of cache usage. The resource allocation unit 231 determines the amount of resource allocation for each combination of a communication port provided in the CM 201 and an HBA of a server connected to that communication port.

The resource allocation unit 231 also detects the processing load (hereinafter sometimes referred to as "I/O load") corresponding to the I/O request for each of the above combinations. Furthermore, the resource allocation unit 231 calculates an estimate of the I/O processing capacity and the response time for the I/O request based on the resource allocation amount and the I/O load. The resource allocation unit 231 registers the I/O processing capacity, I/O load, and response time as performance information in the performance management table 221. Furthermore, the resource allocation unit 231 notifies the server of the performance information registered in the performance management table 221 in response to a request from the server.

CM202 includes a memory unit 220a and a RAID control unit 230a. A performance management table 221a is stored in the memory unit 220a. Similar to the performance management table 221 of CM201, the performance management table 221a registers performance information for each combination of a communication port included in CM202 and an HBA of a server connected to that communication port.

The RAID control unit 230a has the same processing function as the RAID control unit 230 of the CM 201, and executes I/O processing on the data area to be processed in response to an I/O request received from the server. The RAID control unit 230a also includes a resource allocation unit 231a. Like the resource allocation unit 231 of the CM 201, the resource allocation unit 231a allocates resources, detects I/O loads, and calculates response times for each combination of a communication port provided in the CM 202 and an HBA of a server connected to the communication port. The resource allocation unit 231a also registers the resource allocation amount, I/O load, and response time in the performance management table 221a as performance information. The resource allocation unit 231a also notifies the server of the performance information registered in the performance management table 221a in response to a request from the server.

The server 100 comprises a memory unit 110, an I/O control unit 121, and a path control unit 122. The memory unit 110 is realized as a storage area of a storage device provided in the server 100, such as the RAM 102 or the HDD 103. The I/O control unit 121 and the path control unit 122 are realized, for example, by the processor 101 provided in the server 100 executing a predetermined program.

The memory unit 110 stores a performance management table 111. The performance management table 111 holds performance information acquired from the resource allocation units 231, 231a of the CMs 201, 202. The performance information is registered for each combination of a communication port (HBA) of the server 100 and a communication port of the storage device 200 (specifically, a communication port of the CM).

The I/O control unit 121 issues an I/O request to the data area. For example, if the server 100 is a business server, the I/O control unit 121 issues an I/O request in response to a request from a business application. The path control unit 122 selects an access path for transmitting the I/O request issued by the I/O control unit 121, and transmits the I/O request via the selected access path.

For example, if all transmission paths in the active path are alive, the path control unit 122 sends I/O requests equally to those transmission paths. On the other hand, if at least one of the active paths is unavailable, the path control unit 122 determines based on the performance management table 111 whether to send the I/O request using only the remaining transmission paths in the active path, or to send the I/O request using the transmission paths of the standby path as well.

Although not shown, the servers 100a and 100b of the second configuration example also have the same processing functions as the server 100 shown in FIG. 7. Here, the performance management table 111 of the server 100a is registered for each combination of the communication port (HBA) of the server 100a and the communication port of the storage device 200. Also, the performance management table 111 of the server 100b is registered for each combination of the communication port (HBA) of the server 100b and the communication port of the storage device 200.

Figure 8 is a diagram for explaining the resource allocation function of a storage device. As an example, the processing by the resource allocation unit 231 of CM 201 out of CMs 201 and 202 will be explained here.

The resource allocation unit 231 adaptively changes the resource allocation for I/O requests for each communication port (ports #0 to #3) of the CM 201. The RAID control unit 230 executes I/O processing corresponding to the I/O requests received from each communication port according to the amount of resource allocation. This enables I/O performance according to the amount of resource allocation.

In state 1 of FIG. 8, only I/O requests from the active path are input to the communication port of CM 201, and this communication port is used exclusively for the active path. In this state, only I/O requests from the active path need to be processed, so the resources allocated to the active path in this state are set to 100%. In other words, when the communication port is used exclusively for the active path, the resource allocation unit 231 allocates 100% of the resources to the active path.

In addition, in state 2 of FIG. 8, only I/O requests from the standby path are input to the communication port of CM 201, and this communication port is used only for the standby path. In this state, based on the processing capacity allocation amount in state 1, the resource allocation unit 231 allocates 50% of the resources to the standby path. This is because when an I/O request is received from the standby path, inter-CM communication occurs, and the time required for I/O processing is longer than when it is received from the active path. For this reason, it is considered that the remaining 50% of resources will be allocated to inter-CM communication processing.

Note that examples of resources whose allocation is controlled include processor cores and caches. When 50% of resources are allocated, the following specific control is performed compared to when 100% are allocated. For example, 50% of the number of processor cores are allocated to I/O processing in response to a received I/O request. Alternatively, 50% of the capacity of a cache is allocated to that I/O processing. As a result, when the I/O load is the same (the number of I/O requests received per unit time is the same), approximately 50% of the I/O performance is achieved compared to when 100% of resources are allocated.

In state 3 of FIG. 8, both I/O requests from the active path and the standby path are input to the communication port of CM 201, and this communication port is shared by the active path and the standby path. In this state, based on the resource allocation amount in state 1, the resource allocation unit 231 allocates 50% of the resources to the active path and half of that, 25%, to the standby path. As a result, when the I/O load from the active path and the I/O load from the standby path are the same, the I/O processing in response to an I/O request from the active path will perform roughly twice as well as the I/O processing in response to an I/O request from the standby path. Note that the remaining 25% of the resources are considered to be allocated to inter-CM communication processing in response to an I/O request from the standby path.

Based on the resource allocation amount and the detected I/O load value, the resource allocation unit 231 calculates the I/O processing capacity for each transmission path represented by a combination of a communication port of the CM 201 and a connected HBA. In this embodiment, the maximum capacity of one communication port is set to "100", so the capacity is expressed as a percentage.

In state 1 of Figure 8, I/O requests are received only from the active path, so the processing capacity of the transmission path included in the active path is calculated to be 100%. Also, in state 2 of Figure 8, I/O requests are received only from the standby path, so the processing capacity of the transmission path included in the standby path is calculated to be 50%.

On the other hand, in state 3 of FIG. 8, I/O processing is received from both the active path transmission line and the standby path transmission line for one communication port. In this state, the processing capacity of each transmission line is calculated based on the resource allocation amount for each transmission line and the I/O load for each transmission line. For example, if two transmission lines, an active path and a standby path, are connected to a communication port, the processing capacity of each transmission line is calculated by (resource allocation amount for the transmission line) x (ratio of I/O load on the transmission line to the total I/O load on the communication port) x 2.

For example, if the I/O load on the two transmission paths is the same, the processing capacity of each transmission path will be the same as the resource allocation. If the resource allocation is "50%," then 50% x (1/2) x 2 = 50%. Also, if the ratio of the I/O load on the active path transmission path to the I/O load on the standby path transmission path is 1:2, then the processing capacity of the former will be 50% x (1/3) x 2 = approximately 33%, and the processing capacity of the latter will be 25% x (2/3) x 2 = approximately 33%.

In this way, processing capacity indicates the I/O processing volume (number of I/O requests processed) that can be performed using a transmission path, calculated from the resource allocation amount and the detected I/O load. Processing capacity can also be said to be the executable I/O processing volume that is allocated based on the usage status and I/O load of the transmission path.

The process of the resource allocation unit 231 will be described below with reference to a flowchart.
FIG. 9 is an example of a flowchart showing a resource allocation process according to a server connection.
[Step S11] The RAID control unit 230 detects that an HBA of a server has been connected to a communication port.

[Step S12] The resource allocation unit 231 determines whether the detected server is connected via an active path. This determination is made by referencing information (not shown) indicating the correspondence between the access source server, the access destination data area, and the communication ports corresponding to the active path and standby path. If the server is connected via an active path, processing proceeds to step S13; if the server is connected via a standby path, processing proceeds to step S14.

[Step S13] The resource allocation unit 231 allocates 100% of resources to the I/O process in response to an I/O request from the HBA of the detected server.
[Step S14] The resource allocation unit 231 allocates 50% of the resources to I/O processing in response to an I/O request from the HBA of the detected server, and also allocates 50% of the resources to inter-CM communication processing associated with that I/O processing.

The resource allocation amount for each combination of a communication port and a server HBA allocated in step S13 or step S14 above is set in a storage area such as a RAM referenced by the RAID control unit 230 and applied to the RAID control unit 230. Thereafter, the RAID control unit 230 executes I/O processing for received I/O requests based on the processing capacity allocation amount that has been set.

[Step S15] The resource allocation unit 231 registers the resource allocation amount for each combination of a communication port and a server HBA allocated in step S13 or step S14 directly in the corresponding record of the performance management table 221 as the initial processing capacity value for that combination.

Figures 10 and 11 are example flowcharts showing resource allocation processing in response to I/O requests. As shown in Figures 10 and 11, the amount of processing capacity allocated is changed as appropriate depending on the reception status of I/O requests at the communication port.

[Step S21] The RAID control unit 230 detects that a communication port has received an I/O request from a server.
[Step S22] The resource allocation unit 231 determines whether the data area accessed by the I/O request is the area covered by CM 201. If the data area accessed is the area covered by CM 201, the I/O request has been received from the active path, and the process proceeds to step S23. On the other hand, if the data area accessed is not the area covered by CM 201 (in this embodiment, the area covered by CM 202), the I/O request has been received from the standby path, and the process proceeds to step S31 in FIG. 11.

[Step S23] The resource allocation unit 231 determines whether inter-CM communication has occurred in response to an I/O request received at the relevant communication port (the communication port at which the I/O request was received in step S21) during the period from the current time to a time a certain time ago. If inter-CM communication has occurred, the process proceeds to step S25; if inter-CM communication has not occurred, the process proceeds to step S24.

[Step S24] If the answer is No in step S23, the communication port has been used only for the active path for the most recent period of time. Therefore, the resource allocation unit 231 allocates 100% of the resources to the I/O processing in response to the I/O request from the current server (the server that sent the I/O request in step S21).

[Step S25] If step S23 returns Yes, the communication port has been used as both an active path and a standby path for a certain period of time in the past. In this case, the resource allocation unit 231 allocates 25% of the resources to I/O processing in response to an I/O request from another server that uses the communication port as a standby path. The resource allocation unit 231 also allocates 25% of the resources to inter-CM communication processing associated with this I/O processing.

The "other server" in step S25 is a server other than the source server (current server) of the I/O request in step S21. For example, in the case of FIG. 6, if the relevant communication port is port #0 of CM #0 and the current server is server #0, the other server is server #1 and the HBA of the other server is HBA #0 of server #1.

[Step S26] The resource allocation section 231 allocates 50% of the resources to the I/O process in response to the I/O request from the current server.
The resource allocation amount allocated in step S24, steps S25, S26, step S33 described below, or steps S34, S35 described below is overwritten and set in a memory area such as a RAM referenced by the RAID control unit 230, and applied to the RAID control unit 230.

[Step S27] The resource allocation unit 231 detects the I/O load (e.g., the number of I/O requests received in the most recent fixed period of time) at the communication port for each connected HBA. For each combination of the communication port and the connected HBA, the resource allocation unit 231 calculates the processing capacity based on the current resource allocation amount and the detected I/O load. The resource allocation unit 231 also calculates the response time for the I/O request based on the processing capacity and the I/O load. The resource allocation unit 231 overwrites and registers the processing capacity, I/O load, and response time as performance information in the record of the performance management table 221 that corresponds to the combination of the communication port on which the I/O request was received and the connected HBA.

[Step S28] The RAID control unit 230 executes I/O processing for the I/O request based on the updated resource allocation amount.
The explanation will be continued below with reference to FIG.

[Step S31] The resource allocation unit 231 determines whether inter-CM communication has occurred in response to an I/O request received at the communication port in question during the period from the current time to a time a certain amount of time ago. If inter-CM communication has occurred, the processing from step S32 onwards has been executed most recently, and the processing capacity allocation amount from this processing can be applied as is, so processing proceeds to step S27 in FIG. 10. On the other hand, if inter-CM communication has not occurred, processing proceeds to step S32.

[Step S32] The resource allocation unit 231 determines whether an I/O request from another server has been received at the communication port in question during the period from the current time to a time a certain amount of time ago. The "other server" in step S32 is a server other than the source server (current server) of the I/O request in step S21, and is a server that is using the communication port in question as an active path. If an I/O request from another server has been received, processing proceeds to step S34; if an I/O request from another server has not been received, processing proceeds to step S33.

[Step S33] If the result of step S32 is No, the communication port has been used only as a standby path for a certain period of time in the last period, as shown in state 2 of FIG. 8. In this case, the resource allocation unit 231 allocates 50% of the resources to the I/O processing in response to the I/O request from the current server.

[Step S34] If step S32 returns Yes, the communication port has been used as both an active path and a standby path for a certain period of time in the past. In this case, the resource allocation unit 231 allocates 50% of the resources to I/O processing in response to an I/O request from another server that uses the communication port as an active path.

As mentioned in step S32, the "other server" in step S34 is a server that uses the corresponding communication port as an active path. For example, in the case of FIG. 6, if the corresponding communication port is port #0 of CM #0 and the current server is server #1, the other server is server #0 and the HBA of the other server is HBA #0 of server #0.

[Step S35] The resource allocation unit 231 allocates 25% of the resources to the I/O processing in response to the I/O request from the current server. The resource allocation unit 231 also allocates 25% of the resources to the inter-CM communication processing associated with this I/O processing. After this, the process proceeds to step S27 in FIG. 10.

By the above processing of Fig. 10 and Fig. 11, for each communication port of CM201, the allocation of resources to I/O processing in response to I/O requests from the communication port (HBA) on the server side is adaptively controlled according to the usage status of the communication port of CM201. In addition, based on the amount of resource allocation and the detected value of the I/O load, the processing capacity of I/O processing is updated for each transmission path represented by the combination of the communication port and HBA.

Although not shown, the performance management table 221 executes the performance information update process of step S27 at regular time intervals for each communication port of CM 201, regardless of whether an I/O request is received. This causes the performance information in the performance management table 221 to be updated periodically.

In addition, in the description of Figures 8 to 11 above, the processing of the resource allocation unit 231 of CM 201 is explained, but the resource allocation unit 231a of CM 202 also executes the same processing as above. However, the resource allocation unit 231a executes the above processing for each communication port of CM 202.

Figure 12 is a diagram showing an example of the configuration of a CM performance management table. Figure 12 shows an example of the performance management table 221 of CM 201, but the performance management table 221a of CM 202 also has a similar data configuration to that of Figure 12.

As shown in FIG. 12, the performance management table 221 registers the processing capacity, I/O load, and response time for each combination of server WWN (World Wide Name) and port number.

The server WWN is a number for identifying the communication port (HBA) of the connected server. The port number indicates the identification number of the communication port provided in CM 201. In the example of FIG. 12, two server WWNs are associated with the same port number, which indicates, for example, a case in which HBAs of different servers are connected to each communication port as in FIG. 5.

The processing capacity indicates the processing capacity for I/O processing in response to an I/O request from the corresponding HBA, calculated based on the resource allocation amount and the I/O load. The I/O load indicates the detected value of the I/O load due to an I/O request from the corresponding HBA. The response time indicates the estimated value of the response time to an I/O request, calculated based on the processing capacity and the I/O load.

When the resource allocation unit 231 of the CM 201 receives a request for notification of performance information from a server, it notifies the server of the processing capacity, I/O load, and response time registered in the performance management table 221. For example, in FIG. 12, when a communication port with port number "0" receives a notification request from an HBA of server WWN "aaaa", the processing capacity "100%", I/O load "100", and response time "10 ms" associated with server WWN "aaaa" and port number "0" are notified to the HBA of server WWN "aaaa".

Next, the processing on the server side will be described.
First, Fig. 13 is a diagram showing an example of the configuration of a server performance management table. As shown in Fig. 13, the performance management table 111 registers the path type, path status, processing capacity, I/O load, and response time in association with each combination of server WWN and port number. Furthermore, when multiple storage devices are connected to a server, as in the example of Fig. 13, in addition to the server WWN and port number, a storage number that identifies the storage device is also combined.

The server WWN is a number for identifying a communication port (HBA) of the server that stores the performance management table 111. The port number indicates the identification number of the communication port of the connected CM. In the example of FIG. 13, the port numbers "0" to "3" associated with the servers "aaaa", "bbbb", "cccc", and "dddd", and the port numbers "0" to "3" associated with the servers "eeee", "ffff", "gggg", and "hhhh" each indicate the port numbers of different CMs.

The path type indicates whether the access path from the HBA indicated by the server WWN to the communication port of the CM indicated by the port number is being used as an active path or a standby path. The path status indicates whether the access path is available (Enable) or unavailable (Disable).

In each of the processing capacity, I/O load, and response time fields, information notified from the CM via the corresponding access path is registered. Processing capacity indicates the processing capacity for I/O processing in response to an I/O request via the corresponding access path. I/O load indicates the detected value of the I/O load due to an I/O request via the corresponding access path. Response time indicates an estimated value of the response time to an I/O request, calculated based on processing capacity and I/O load.

When the server connects to a CM, and at regular time intervals thereafter, the path control unit 122 of the server outputs a request for notification of performance information for each communication port of the connected CM. The path control unit 122 overwrites and registers the performance information notified in response to the notification request in the performance management table 111.

<First Configuration Example>
Next, the processing of the server 100 in the first configuration example will be described.
Fig. 14 is a diagram showing an example of a normal state. In Fig. 14, all transmission paths in the active path are in a state where communication is possible normally. In this case, the path control unit 122 of the server 100 distributes I/O requests to the data area 211 evenly to each transmission path in the active path and transmits them.

Table 111a in FIG. 14 shows performance information extracted from the information in performance management table 111 held by server 100. As described above, the active path includes transmission paths from HBAs #0 to #3 of server 100 to ports #0 to #3 of CM #0, respectively, and the standby path includes transmission paths from HBAs #4 to #7 of server 100 to ports #0 to #3 of CM #1, respectively. In the initial state, 100% of the processing capacity is assigned to each transmission path of the active path, and 50% of the processing capacity is assigned to each transmission path of the standby path.

In the example of FIG. 14, when 100% processing capacity is assigned to each transmission line of the active path, and the index indicating the I/O load (e.g., the number of I/O requests issued per unit time) is "100", the response time to an I/O request will be 10 ms. 10 ms is the upper limit of the response time (maximum performance).

Figure 15 is a diagram showing a first example of a state in which a failure has occurred in an active path. In this figure, as an example, let us assume that, in the state of Figure 14, failures have occurred in two transmission paths from HBA #0 and #1 of server 100 to ports #0 and #1 of CM #0, respectively, and these transmission paths have become unusable. In this case, based on performance information before the failure occurred, the path control unit 122 of server 100 estimates I/O performance (response time) for both cases in which an I/O request is sent using only the remaining surviving transmission paths in the active path, and in which an I/O request is sent using the transmission paths in the standby path in addition to the remaining surviving transmission paths.

Table 111b1 in FIG. 15 shows estimated values of performance information when only the remaining transmission paths in the active path are used. In this case, the path control unit 122 distributes I/O requests that were previously sent using the four transmission paths in the active path evenly to the two transmission paths from HBA #2 and #3. In this case, the I/O load on each of the two remaining transmission paths doubles, and it is estimated that the response time will also double. In other words, the I/O load index for each transmission path will be "200", and the response time will be 20 ms.

On the other hand, table 111b2 in FIG. 15 shows estimated values of performance information when the standby path transmission path is used in addition to the remaining transmission paths in the active path. In this case, the path control unit 122 distributes I/O requests to the remaining transmission paths in the active path and the standby path transmission paths so that the I/O performance (response time) is equalized.

Since each transmission line of the standby path is assigned 50% of the processing capacity, it is estimated that the response time of each transmission line will be equalized if the I/O load (number of I/O requests issued) on each transmission line of the standby path is set to half that of each transmission line of the active path. This condition is met by setting the I/O load of each transmission line of the active path to "100" and the I/O load of each transmission line of the standby path to "50", as shown in table 111b2. In this case, it is estimated that the response time on each transmission line will be 10 ms.

Here, assume that the upper threshold (tolerable limit) of the response time on the active path is set to 20 ms. In this case, even if only the remaining transmission paths of the active path are used, the response time will not exceed the upper threshold. For this reason, the path control unit 122 determines that there is no need to use the standby path, and distributes I/O requests to the data area 211 evenly across the remaining transmission paths of the active path.

Figure 16 is a diagram showing a second example of a state in which a failure has occurred in an active path. In this example, assume that in the state shown in Figure 15 where communication is performed using only the remaining transmission paths in the active path, a failure occurs in the transmission path from HBA #2 of server 100 to port #2 of CM #0, causing this transmission path to become unusable. In this case, based on performance information before the failure, the path control unit 122 of server 100 estimates I/O performance (response time) for both cases in which an I/O request is sent using only the remaining surviving transmission paths in the active path, and in which an I/O request is sent using the transmission paths in the standby path in addition to the remaining surviving transmission paths.

Table 111c1 in FIG. 16 shows estimated values of performance information when only the remaining transmission paths in the active path are used. In this case, the path control unit 122 transmits an I/O request that was previously transmitted using two transmission paths in the active path using only one transmission path from HBA #3. In this case, the I/O load on the remaining transmission path is doubled compared to the state of table 111b1 in FIG. 15, so it is estimated that the response time will also double. In other words, the I/O load index for each transmission path is "400", and the response time is 40 ms, exceeding the upper threshold of 20 ms.

On the other hand, table 111c2 in FIG. 16 shows estimated values of performance information when the standby path transmission path is used in addition to the remaining transmission paths in the active path. In this case, the path control unit 122 distributes I/O requests to the remaining transmission paths in the active path and the standby path transmission paths so that the I/O performance (response time) is equalized.

As mentioned above, each transmission line of the standby path is assigned 50% of the processing capacity, so if the I/O load on each transmission line of the standby path is set to half that of each transmission line of the active path, it is estimated that the response time of each transmission line will be equal. As shown in table 111c2, this condition is met by setting the I/O load of the transmission line of the active path to "133" and the I/O load of each transmission line of the standby path to "67". In this case, it is estimated that the response time for each transmission line will be 13 ms.

In this way, the response time on the transmission path of the active path is shortened compared to when only the remaining transmission path of the active path is used, so the path control unit 122 determines that it is better to use the standby path. The path control unit 122 distributes and transmits I/O requests to the data area 211 so that the I/O load on the remaining transmission path of the active path is "133" and the I/O load on each transmission path of the standby path is "67".

Figure 17 shows a third example of a state in which a failure has occurred in an active path. In Figures 14 to 16 above, a relatively large I/O load of "400" was applied to all access paths to the data area 211. In such a case, when many transmission paths in the active path become unusable as shown in Figure 16, it is determined that there is a high need to also use a standby path. However, when the I/O load of the entire access path is smaller, even if many transmission paths in the active path become unusable, it may be possible to maintain I/O performance at or above the allowable value by using only the remaining transmission paths of the active path without using the standby path.

As shown in table 111d1 in FIG. 17, when all transmission paths in the active path are normal, the I/O load of each transmission path is "25". In this state, the response time of each transmission path is 10 ms, which corresponds to the maximum performance. From this state, it is assumed that a failure occurs in three transmission paths that reach ports #0-#2 of CM#0 from HBA#0-#2 of server 100, respectively, and these transmission paths become unusable. In this case, the path control unit 122 of server 100 estimates the I/O performance (response time) for both cases, based on the performance information before the failure, when an I/O request is sent using only the remaining surviving transmission path in the active path, and when an I/O request is sent using the transmission path in the standby path in addition to the one remaining surviving transmission path.

Table 111d2 in FIG. 17 shows estimated values of performance information when only the one remaining transmission path in the active path is used. In this case, the path control unit 122 transmits I/O requests that were previously sent using the four transmission paths in the active path using only the one transmission path from HBA#3. In this case, the I/O load on the remaining transmission path becomes four times larger, or "100," than in the state of table 111d1, but in this state the response time is maintained at 10 ms and does not exceed the upper threshold of 20 ms. For this reason, the path control unit 122 determines that it is not necessary to use the standby path, and transmits I/O requests to the data area 211 using only the one remaining transmission path of the active path.

As described above, when the transmission path of the active path becomes unavailable, the server can effectively use the standby path and suppress the degradation of I/O performance by using not only the remaining transmission paths in the active path but also the standby path. Here, the case where the standby path is used is limited to the case where the I/O load of each transmission path exceeds the upper threshold when only the remaining transmission paths in the active path are used. Therefore, the standby path, which is a spare access path, can be used only when it is highly necessary to send an I/O request. In addition, the I/O load of the transmission path is estimated for both the method of using only the remaining transmission paths in the active path and the method of using the standby path as well, and the method that reduces the I/O load is selected. As a result, the standby path is used only when it is possible to improve I/O performance, and unnecessary use of the standby path can be avoided.

This type of processing improves the efficiency of transmission path utilization and makes it possible to maintain a balanced I/O performance for the entire system. This effect can be achieved simply by modifying the software used to control path selection, without changing the access path configuration.

<Second Configuration Example>
Next, the processing of the servers #0 and #1 (servers 100a and 100b) in the second configuration example will be described.

Figure 18 is a diagram showing an example of a normal state. In Figure 18, all transmission paths in the active paths for both servers #0 and #1 are in a state where communication is possible normally. In this case, the path control unit 122 of server #0 distributes I/O requests for data area 212 evenly to each transmission path in the corresponding active path and sends them. Also, the path control unit 122 of server #1 distributes I/O requests for data area 213 evenly to each transmission path in the corresponding active path and sends them.

Table 111e in FIG. 18 shows performance information extracted from the information in the performance management table 111 held by servers #0 and #1. As described above, the active path of server #0 includes the transmission paths from HBA #0 and #1 of server #0 to ports #0 and #1 of CM #0. The standby path of server #0 includes the transmission paths from HBA #2 and #3 of server #0 to ports #0 and #1 of CM #1. On the other hand, the active path of server #1 includes the transmission paths from HBA #0 and #1 of server #1 to ports #0 and #1 of CM #1. The standby path of server #1 includes the transmission paths from HBA #2 and #3 of server #1 to ports #0 and #1 of CM #0. In the initial state, 100% of the processing capacity is assigned to each transmission path of the active path, and 50% of the processing capacity is assigned to each transmission path of the standby path.

In the example of FIG. 18, when 100% of the processing capacity is assigned to each transmission line of the active path and the index indicating the I/O load is "100", the response time to an I/O request will be 10 ms. 10 ms is the upper limit of the response time (maximum performance).

Figure 19 is a diagram showing a fourth example of a state in which a failure has occurred in an active path. In this example, in the state of Figure 19, let us assume that a failure has occurred in the transmission path from HBA #0 of server #0 to port #0 of CM #0 in the state of Figure 18, causing this transmission path to become unusable. In this case, the path control unit 122 of server #0 estimates the I/O performance (response time) based on the performance information before the failure occurred for both cases in which an I/O request is sent using only the one remaining surviving transmission path in the active path, and in which an I/O request is sent using the transmission path in the standby path in addition to the one remaining surviving transmission path.

Table 111f in Figure 19 shows the estimated performance information when only the one remaining transmission path in the active path of server #0 is used. In this case, the I/O load on the remaining transmission path doubles, so it is estimated that the response time will also double. In other words, the I/O load index for the remaining transmission path will be "200", and the response time will be 20 ms.

Let us now assume that the upper threshold (tolerable limit) of the response time on the active path of server #0 is set to 20 ms. In this case, as shown in table 111f, the response time will not exceed the upper threshold even if only the remaining transmission path of the active path is used. For this reason, the path control unit 122 of server #0 determines that there is no need to use the standby path, and sends an I/O request for the data area 212 to the one remaining transmission path of the active path.

Figure 20 is a diagram showing a fifth example of a state in which a failure has occurred in an active path. In this example, assume that in Figure 19, communication was performed using only the one remaining transmission path in the active path of server #0, and then a failure occurred in the transmission path from HBA #1 of server #0 to port #1 of CM #0, rendering this transmission path unusable. In this case, since all transmission paths in the active path are unusable, the path control unit 122 of server #0 distributes I/O requests to the data area 212 evenly to each transmission path of the standby paths and sends them. At this time, the I/O load of each transmission path of the standby paths is "100".

As a result, ports #0 and #1 of CM #1 are used not only for the active path of server #1, but also for the standby path of server #0. This causes the allocation of processing power by the resource allocation unit 231a of CM #1 to change as follows. That is, the allocation changes from a state in which 100% of processing power is allocated to the active path of server #1 to a state in which 50% of processing power is allocated to the active path of server #1 and 25% of processing power is allocated to the standby path of server #0.

As a result of this change in allocation amount, as shown in table 111g in FIG. 20, the response time for each transmission line in the standby path of server #0 becomes 80 ms. Also, the response time for each transmission line in the active path of server #1 becomes 20 ms, resulting in a decrease in I/O performance.

By periodically acquiring performance information from CM #1, the path control unit 122 of server #1 recognizes that the amount of processing capacity allocated to each transmission line of the active path has become smaller as described above, causing a decline in I/O performance. Here, assume that the upper threshold (tolerable limit value) of the response time for the active path of server #1 is set to 15 ms. In this case, the reduction in the amount of processing capacity allocated causes the response time for each transmission line in the active path to exceed the upper threshold. When the path control unit 122 of server #1 determines this, it estimates the I/O performance (response time) when an I/O request is sent using the transmission lines in the standby path in addition to the active path, based on the acquired changed performance information.

Figure 21 shows estimated values of performance information when active paths and standby paths are used in combination. When standby paths are also used in combination in response to periodic acquisition of performance information from the CM, the path control unit 122 of server #1 distributes I/O requests to each transmission path of the standby paths up to the upper limit of their processing capabilities.

As shown in table 111h1 in FIG. 21, 50% of the processing capacity is assigned to each transmission line of the standby path, so if I/O requests are distributed so that the I/O load of each transmission line is "50", it is estimated that the response time on each transmission line will be the shortest, 10 ms. At this time, the I/O load on each transmission line of the active path can be reduced to "50", thereby shortening the response time on each transmission line to 10 ms, which is below the upper threshold. Therefore, the path control unit 122 of server #1 determines that it is better to also use the standby path, and distributes and sends I/O requests to the data area 213 so that the I/O load on each transmission line of the active path and standby path is "50".

As in the second configuration example above, when the same communication port of a CM is used for both the active path and the standby path, even if all transmission paths of the active path are alive, the I/O performance of the active path may fall below the allowable value depending on the usage status of the standby path. As shown in table 111h1, even if all active paths are alive, if the I/O performance falls below the allowable value, the server can recover the I/O performance by sending an I/O request using the standby path as well.

When the resource allocation unit 231a of CM #1 detects that the I/O load on the active path of server #1 has decreased, it recalculates the allocation ratio between the processing capacity of each transmission line of this active path and the processing capacity of each transmission line of the standby path of server #0, which uses the same communication port. As a result of this recalculation, as shown in table 111h2 in FIG. 21, 33% of the processing capacity is allocated to each transmission line of the active path of server #1, and 33% of the processing capacity is allocated to each transmission line of the standby path of server #0. As a result, the response time on the standby path of server #0 is reduced to 60 ms, and I/O performance is improved.

In this way, reducing the I/O load on the active path of server #1 in order to recover the I/O performance on that path may result in improved I/O performance on the standby path of server #0, which uses the same communication path on the CM side. In other words, by constantly monitoring I/O performance, not just when the transmission path becomes unavailable, and sending I/O requests using the standby path if there is a significant drop in performance, the utilization efficiency of the transmission path is improved, making it possible to maintain a balanced I/O performance for the entire system.

In the state of table 111h2, the response time on the active path of server #1 deteriorates to 15 ms. In this example, the response time is below the upper threshold, but if the response time on that path exceeds the upper threshold, the path control unit 122 of server #1 redistributes the I/O load based on the updated processing capacity. In this way, the I/O performance on each transmission path is optimized as appropriate.

Next, the process of the server will be described with reference to a flowchart.
FIG. 22 is an example of a flowchart showing a path selection process when a failure occurs.
[Step S41] The path control unit 122 of the server detects the occurrence of a fault in the transmission path connected to the HBA. Then, the process proceeds to step S42 and subsequent steps.

[Step S42] The path control unit 122 determines whether any of the transmission paths included in the active path are alive (enabled). If any of the transmission paths are alive, the process proceeds to step S43. If no of the transmission paths are alive, the process proceeds to step S48.

[Step S43] The path control unit 122 refers to the performance management table 111 and calculates the response time Ta of each transmission path when I/O requests are evenly distributed to the surviving transmission paths in the active path (when the I/O load is evenly distributed).

[Step S44] The path control unit 122 determines whether the calculated response time Ta is equal to or less than a predetermined upper threshold. If the response time Ta is equal to or less than the upper threshold, the process proceeds to step S47. If the response time Ta exceeds the upper threshold, the process proceeds to step S45.

[Step S45] The path control unit 122 refers to the performance management table 111 and calculates the response time Tb of each transmission path when I/O requests are distributed so that the response times of the surviving transmission paths in the active path and the surviving transmission paths in the standby path are equal.

[Step S46] The path control unit 122 determines whether the calculated response time Tb is equal to or less than the response time Ta. If the response time Tb is equal to or less than the response time Ta, the process proceeds to step S47. If the response time Tb is longer than the response time Ta, the process proceeds to step S48.

[Step S47] The path control unit 122 distributes I/O requests evenly among the surviving transmission paths in the active path.
[Step S48] The path control unit 122 distributes I/O requests between surviving transmission lines in the active path and surviving transmission lines in the standby path so that the response times of the respective transmission lines are equal.

23 is an example of a flowchart showing a path selection process when performance information is acquired. The process in FIG. 23 is repeatedly executed at regular time intervals.
[Step S51] The path control unit 122 of the server requests the CM connected to each transmission path to notify it of performance information about each transmission path of the active path and the standby path corresponding to one data area. In response to this request, the CM notifies it of the performance information, and when the path control unit 122 acquires this performance information, it overwrites it in the performance management table 111 and registers it.

[Step S52] The path control unit 122 determines whether any of the transmission paths included in the active path are surviving. If any of the transmission paths are surviving, the process proceeds to step S53. If no surviving transmission paths are surviving, the process in FIG. 23 ends.

[Step S53] The path control unit 122 determines whether the I/O load of each surviving transmission path in the active path is excessive. Specifically, if the I/O load of the transmission path registered in the performance management table 111 exceeds a predetermined upper threshold, the I/O load is determined to be excessive. This upper threshold may be the same as the value used in step S44 in FIG. 22. If the I/O load is excessive, processing proceeds to step S54; if the I/O load is not excessive, the processing in FIG. 23 ends.

[Step S54] The path control unit 122 distributes I/O requests to the surviving transmission paths in the active path and the surviving transmission paths in the standby path. At this time, the path control unit 122 adjusts the distribution ratio so that I/O requests are distributed to each transmission path of the standby path up to the upper limit of the processing capacity.

As another example, in step S54, similar to step S48 in FIG. 22, I/O requests may be distributed to the surviving transmission paths in the active path and the surviving transmission paths in the standby path so that the response times for each transmission path are equal.

The processing functions of the devices (e.g., information processing device 10, servers 100, 100a, 100b, CMs 201, 202) shown in each of the above embodiments can be realized by a computer. In this case, a program describing the processing contents of the functions that each device should have is provided, and the above processing functions are realized on the computer by executing the program on a computer. The program describing the processing contents can be recorded on a computer-readable recording medium. Examples of computer-readable recording media include magnetic storage devices, optical disks, and semiconductor memories. Examples of magnetic storage devices include hard disk drives (HDDs) and magnetic tapes. Examples of optical disks include CDs (Compact Discs), DVDs (Digital Versatile Discs), and Blu-ray Discs (BD, registered trademark).

When distributing a program, for example, portable recording media such as DVDs and CDs on which the program is recorded are sold. The program can also be stored in a storage device of a server computer, and the program can be transferred from the server computer to other computers via a network.

A computer that executes a program stores, for example, a program recorded on a portable recording medium or a program transferred from a server computer in its own storage device. The computer then reads the program from its own storage device and executes processing according to the program. Note that the computer can also read a program directly from a portable recording medium and execute processing according to that program. The computer can also execute processing according to the received program each time a program is transferred from a server computer connected via a network.

1a, 1b, 2a, 2b Path 3a Communication load 3b Processing capacity 10 Information processing device 11 Communication unit 12 Processing unit 20 Storage device 21 Storage area P1 to P4, P11 to P14 Path

Claims (7)

a communication unit that is connected to the storage device via a plurality of first paths of an active system and a plurality of second paths of a standby system;
when some of the plurality of first paths become unavailable, by referencing a communication load for each of the remaining first paths and the plurality of second paths that are available among the plurality of first paths and a processing capacity allocated by the storage device , a first response time when an access request to the storage device is distributed to the remaining first paths is estimated based on the communication load and the processing capacity for each of the remaining first paths, and a second response time when an access request to the storage device is distributed to the remaining first path and the plurality of second paths is estimated based on the communication load and the processing capacity for each of the remaining first paths and the plurality of second paths;
a processing unit that determines whether to distribute an access request to the storage device to the remaining first paths or to distribute an access request to the storage device to the remaining first paths and the plurality of second paths based on a comparison result between the first response time and the second response time;
An information processing device having the above configuration. The processing unit includes:
If the first response time is equal to or less than a predetermined first threshold, distribute the access requests to the storage device to the remaining first paths;
if the first response time exceeds the first threshold, based on a comparison result between the first response time and the second response time, determine whether to distribute the access requests to the storage device to the remaining first paths or to distribute the access requests to the storage device to the remaining first paths and the plurality of second paths;
2. The information processing device according to claim 1 . the processing capabilities for each of the remaining first paths and the plurality of second paths are allocated by the storage device in accordance with a usage status of each communication port on the storage device side of each of the remaining first paths and the plurality of second paths as an active path and a standby path, and the current communication load on each communication port;
3. The information processing device according to claim 1 or 2 . the storage device has a storage area to be accessed by the information processing device, and a first control device and a second control device that control access to the storage area;
when an access request to the storage area is transmitted to the plurality of first paths, the storage area is accessed via the first control device, and when an access request to the storage area is transmitted to the plurality of second paths, the storage area is accessed via the second control device and the first control device.
The information processing device according to claim 1 . The processing unit includes:
acquiring the processing capabilities for each of the plurality of first paths and the plurality of second paths from the storage device at regular time intervals, and estimating a third response time when an access request to the storage device is distributed to the plurality of first paths based on the newly acquired processing capabilities for the plurality of first paths and the current communication loads for the plurality of first paths;
when the third response time exceeds a predetermined second threshold, distributing an access request to the storage device among the plurality of first paths and the plurality of second paths;
3. The information processing device according to claim 1 or 2 . a communication port on the storage device side in the plurality of first paths is also used as a standby path for accessing the storage device from another information processing device,
the processing capabilities of the plurality of first paths are allocated by the storage device in accordance with a usage status of an active path from the information processing device and a standby path from the other information processing device at the communication port, and a load of access requests from the information processing device and the other information processing device at the communication port;
6. The information processing device according to claim 5 . The computer
in a state in which a storage device is connected via a plurality of first paths of an active system and a plurality of second paths of a standby system, when some of the plurality of first paths become unavailable, by referring to a communication load for each of the remaining first paths and the plurality of second paths that are available among the plurality of first paths and a processing capacity allocated by the storage device, a first response time when an access request to the storage device is distributed to the remaining first paths is estimated based on the communication load and the processing capacity for each of the remaining first paths, and a second response time when an access request to the storage device is distributed to the remaining first paths and the plurality of second paths is estimated based on the communication load and the processing capacity for each of the remaining first paths and the plurality of second paths;
determining whether to distribute the access requests to the storage device to the remaining first paths or to distribute the access requests to the storage device to the remaining first paths and the plurality of second paths based on a comparison result between the first response time and the second response time;
Path control method.

Images