JP2005010956A - Method for controlling information processor, information processor, information processing system and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method for controlling an information processor by which more appropriate load distribution can be realized even though the load distribution between input/output paths may not be necessarily attained in a round robin system for sequentially assigning data input/output requests to the input/output paths without considering the respective configurations or conditions of the input/output paths. <P>SOLUTION: In this method for controlling an information processor, which transmits a data input/output request through a plurality of logical paths set between storage devices communicatively connected to each other to the storage devices, the loads of the logical paths are acquired, and the logical paths for assigning data input/output requests are decided according to the loads. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a method for controlling an information processing apparatus, an information processing apparatus, an information processing system, and a program.
[0002]
[Prior art]
In recent years, the amount of data handled by computer systems has increased, and the capacity of storage devices has been increasing. In a large-capacity storage device, it is necessary to ensure input / output performance corresponding to the storage capacity. There is also a need to improve system reliability. For this reason, a technique for multiplexing (redundant) input / output paths to the storage device has been developed.
[0003]
When the input / output paths to the storage device are multiplexed, it is necessary for the computer to perform processing for distributing data input / output requests to the storage apparatus to the respective input / output paths. In the conventional technique, in order to distribute the load applied to each input / output path, data input / output requests are equally distributed to each input / output path. As a distribution method, for example, a round robin method in which each input / output path is specified in order is adopted.
[0004]
[Non-Patent Document 1]
W. By Curtis Preston, translated by Hiroki Kanesaki, translated by Toyosawa Jun
“SAN & NAS Storage Network Management” O'Reilly Japan,
October 30, 2002, p. 66-67
[0005]
[Problems to be solved by the invention]
By the way, the input / output path set by multiplexing is not necessarily configured by hardware or software having the same performance, and all or part of the same input / output path is formed by a plurality of information processing apparatuses. It may be shared. That is, the load on each input / output path varies depending on various factors. Therefore, in the round robin method in which data input / output requests are assigned to the input / output paths in order without considering the individual configuration or circumstances of the input / output paths, the load distribution between the input / output paths is not necessarily achieved properly. There is.
[0006]
The present invention has been made from such a viewpoint, and has as its main object to provide a method for controlling an information processing apparatus, an information processing apparatus, an information processing system, and a program that enable more appropriate load distribution. .
[0007]
[Means for Solving the Problems]
A main invention of the present invention is an information processing device control method for transmitting a data input / output request to a storage device through a plurality of logical paths set between the storage devices communicably connected. The load is acquired, and the logical path to which the data input / output request is assigned is determined according to the load.
[0008]
As described above, according to the above-described invention, since data input / output requests are allocated to logical paths according to the logical path load, the load is less concentrated on a specific logical path (bottleneck elimination). Thus, load distribution can be effectively achieved as compared with the conventional round robin method. In addition, a logical path may be set so that a route on the way is shared by a plurality of information processing apparatuses. However, in the present invention, allocation of a data input / output request to a logical path is activated according to the load on the logical path. Therefore, load distribution is appropriately performed in consideration of fluctuations in the load of the logical path by other information processing apparatuses, and load distribution is effectively achieved.
[0009]
Note that other features of the present invention will become apparent from the description of the present specification and the accompanying drawings.
[0010]
DETAILED DESCRIPTION OF THE INVENTION
=== Overall Configuration of Information Processing System ===
FIG. 1 shows the overall configuration of an information processing system (computer system) according to the present embodiment. Two information processing devices (hosts) are connected to one storage device (storage device, storage, disk array device). The information processing apparatus 100 includes a CPU (Central Processing Unit) and a memory, and executes various application programs. For example, programs such as a bank automatic deposit and withdrawal system and an aircraft seat reservation system are executed.
[0011]
The information processing apparatus 100 accesses data stored in the storage device 200 when executing the application program. The data access is performed using a logical path associated with a physical path that is a data transfer path physically configured by hardware connecting the information processing apparatus 100 and the storage apparatus 200 as a transfer path. Connection units 300 (301 to 308) that connect the information processing apparatus 100 and the storage device 200 configure the physical path. The IF (Inter Face) 140 (141 to 148) is, for example, an HBA (Host Bus Adapter) having one or more HP Ports (Host bus adapter Ports).
[0012]
The storage device 200 includes a logical unit 220 and a disk control unit 210. The logical unit 220 is a storage area that is logically set in a storage resource provided to the information processing apparatus 100. In FIG. 1, two logical units of LU (Logical Unit) 1 (221) and LU2 (222) are shown. Various storage resources such as a hard disk device, a flexible disk device, and a semiconductor storage device can be used as the storage resource. The disk control unit 210 receives a data input / output request from the information processing apparatus 100 and controls access to data stored in the logical unit 220. Here, the data input / output request is, for example, a SCSI command. The SCSI command includes, for example, a command for writing data to the disk drive (hereinafter referred to as a data write command) and a command for reading data from the disk drive (hereinafter referred to as a data read command). As shown in FIG. 1, the disk control unit 210 can be integrated with the storage device 200, or can be a separate type. In FIG. 1, two logical units are shown. However, one logical unit 220 may be divided into a plurality of partitions. A partition is a storage area configured by logically dividing the logical unit 220.
[0013]
In the information processing apparatus 100, various application programs 110 are executed. Both application programs 110 execute processing while appropriately accessing data stored in the storage device 200. An accessible logical unit 220 is assigned to the application program 110.
[0014]
=== About Logical Path ===
A data input / output request request from the application program 110 to the storage device 200 is transmitted using a logical path as a transfer path under the control of a path management driver 120 described later. Here, the logical path is defined by a combination of the information processing apparatus 100, the storage apparatus 200, and one or more pieces of software or one or more pieces of hardware constituting a communication path connecting the information processing apparatus 100 and the storage apparatus 200. .
[0015]
FIG. 2 shows a specific example of elements for defining a logical path. As shown in this figure, the logical paths are HLU (Host Logical Unit) 223, TID (Target ID) 181, Bus 182, HP Port (Host bus adapter Port) 180, DPort (Disk controller Port) 230, disk controller 210, and LU. It is defined by specifying hardware and software. HLU 223, TID 181, Bus 182, and HP Port 180 are concepts used in the SCSI (Small Computer Systems Interface) communication standard, and indicate a host logical unit, a target ID, a logical bus, and a host bus adapter side port, respectively. DPort 230 indicates a port of the disk control unit 210. The definition of the logical path does not necessarily depend on the above elements, and any one of the above elements can be made unnecessary by the system, or the logical path can be defined by adding other elements.
[0016]
=== Path management driver ===
The path management driver 120 is software for multiplexing the data transfer path between the information processing apparatus 100 and the storage device 200. The path management driver 120 connects the information processing apparatus 100 and the storage apparatus 200 with a plurality of logical paths 300 and multiplexes data transfer paths, thereby eliminating data access bottlenecks and improving data input / output processing performance. Improve. The path management driver 120 also performs load distribution by appropriately distributing data input / output requests to each input / output path. The path management driver 120 also has a function of blocking the data transfer path when a failure occurs in the data transfer path and continuing data access to the storage device 200 through the normal data transfer path. Service stoppage due to transfer path failure can be avoided and system reliability can be improved.
[0017]
Note that some operating systems executed by the information processing apparatus 100 do not take into account that the storage apparatus 200 is connected by a plurality of logical paths. When such an operating system is executed by the information processing apparatus 100, if the information processing apparatus 100 and the storage device 200 are connected by a plurality of logical paths, the application program 110 exists on the storage device 200. There arises a problem that it appears as if one logical unit 220 exists for each logical path. In order to avoid this, the path management driver 120 also has a function of integrating the original one logical unit 220 that can be seen for each logical path into one and making it appear to the application program 110.
[0018]
=== Logical path assignment processing ===
Processing for assigning data input / output requests to logical paths will be described. This process is performed by the path management driver 120. The path management driver 120 assigns a data input / output request input from the application program 110 to any one of a plurality of logical paths. The path management driver 120 can perform this allocation by a general round robin method, and can perform it according to the load of each logical path. In the latter method, the load on each logical path is grasped as, for example, the number of data input / output requests that pass through the logical path per unit time or the amount of data that passes through the logical path per unit time. Note that an operator or the like sets in the path management driver 120 which of the above allocation methods is adopted. Alternatively, the path management driver 120 can automatically select the selection.
[0019]
The path management driver 120 acquires the load of each logical path at an appropriate timing. The load on the logical path is acquired as, for example, the time from when the path management driver 120 transmits a data input / output request to the logical path until the response to the data input / output request is received.
[0020]
In the round robin method, data input / output requests are assigned in a predetermined order regardless of the load of the logical path. For example, even if there is another low-load logical path, A new data input / output request may be assigned to the path. On the other hand, in the method of assigning data input / output requests to logical paths according to the load, data input / output requests are assigned so that each logical path is efficiently used as a whole. For example, a new data input / output request is assigned to the logical path with the smallest load at the time of assignment, and control is performed so that the load is not concentrated on a specific logical path.
[0021]
=== Get load of logical path ===
Next, a mechanism in which the path management driver 120 determines a logical path to which a data input / output request is assigned according to the load on the logical path will be described in detail.
First, a mechanism in which the path management driver 120 acquires the load of each logical path will be described. FIG. 3 is a flowchart illustrating an example of processing for acquiring the load of each logical path. In this example, a data input / output request is transmitted from the path management driver 120 to the logical path, and a time until the path management driver 120 receives a response to the data input / output request (hereinafter referred to as response time) is measured. By doing so, the load of each logical path is acquired. That is, the logical path with a longer response time has a larger load, and the shorter the response time, the smaller the load.
[0022]
In the process shown in FIG. 3, the response time of each logical path is measured regularly at regular time intervals. When the timing for measuring the response time comes (S3001: YES), the path management driver 120 transmits a data input / output request (hereinafter referred to as a test request) to each logical path (S3002). Here, a data input / output request (for example, a command for writing data having the same data size when the data input / output request is a data write command) is transmitted to each logical path. The path management driver 120 stores the time when the test request is transmitted to the logical path.
[0023]
When the path management driver 120 receives a response to the test request transmitted to each logical path (S3003), a time interval between the time when the test request corresponding to the response is transmitted and the time when the response is received (that is, Response time) is calculated (S3004). The response time measured for each logical path is registered in the logical path management table 400 shown in FIG. 4 (S3005). Each logical path is given an identifier (hereinafter referred to as a logical path ID) that is uniquely set. The logical path management table 400 is referred to by the path management driver 120. The logical path management table 400 is stored in, for example, a memory or a hard disk device of the information processing apparatus 100.
[0024]
=== Determination of allocation priority ===
Next, the allocation priority determined based on the response time of each logical path registered in the logical path management table 400 as described above will be described. The allocation priority is data that is referred to by the path management driver 120 when determining which logical path the data input / output request is allocated to. In this embodiment, the allocation priority is determined to be a larger value for a logical path with a smaller load. In other words, the allocation priority is determined to be larger as the measured logical path has a shorter response time. Specifically, with reference to a logical path with a large load, the allocation priority is determined so as to reduce the load on the logical path. Hereinafter, a process in which the path management driver 120 determines the allocation priority based on the response time and sets the content of the allocation priority in the logical path management table 400 will be specifically described.
[0025]
FIG. 5 is a flowchart for explaining a process in which the path management driver 120 determines the allocation priority and sets the allocation priority in the logical path management table 400 (hereinafter referred to as an allocation priority setting process). The allocation priority setting process is started at an appropriate timing by the path management driver 120 (S5001). For example, it is activated in conjunction with the response time of the logical path management table 400 being updated. Also, for example, it is activated when a timing set in advance for the path management driver 120 by an operator or the like has arrived.
[0026]
The assignment priority is set for each group of logical paths connected to one logical unit (LU). That is, in the assignment priority setting process, load distribution is realized within a group of logical paths connected to one logical unit. Hereinafter, a group consisting of a group of logical paths that realize load sharing with each other will be referred to as a logical path group. In carrying out the present invention, it is not always necessary to realize load distribution within the range of the group of logical paths. The group of logical paths can be set in an arbitrary range, for example, targeting logical paths connected to a plurality of logical units.
[0027]
When the allocation priority setting process is activated, a theoretical load is calculated for a group of logical paths connected to a certain logical unit according to a predetermined algorithm using the allocation priority and response time as parameters (S5002). The theoretical load is an intermediate parameter used in determining the allocation priority. As the allocation priority of each logical path, a value already registered as the content of the allocation priority in the logical path management table 400 is referred to. In an initial state such as when the information processing apparatus 100 is activated or reset, a default value is set for the allocation priority in the logical path management table 400. As the default value, for example, an intermediate value of the allocation priority range is adopted. Further, for example, the default value is set at a ratio corresponding to the measured response time. The default value can be arbitrarily set by the user.
[0028]
As the predetermined algorithm, for example, an algorithm having a theoretical load that is a result of four arithmetic operations (for example, simple product) using allocation priority and response time as parameters is employed. It is assumed that the predetermined algorithm increases the theoretical load as the logical path load increases and as the logical path allocation priority increases.
When the theoretical load is calculated, the path management driver 120 selects a logical path having the maximum theoretical load among the calculated theoretical loads of each logical path as a reference logical path (S5003).
[0029]
Next, the path management driver 120 adds a predetermined value (hereinafter, changed value) to the content of the allocation priority of the logical path management table 400 for one of the logical paths other than the reference logical path, and the value after the addition And the value set in the response time of the logical path management table 400 as parameters, the theoretical load (hereinafter referred to as comparative theoretical load) is calculated again by the predetermined algorithm or other algorithms (S5004).
[0030]
Next, the path management driver 120 compares the calculated comparative theoretical load with the theoretical load calculated for the reference logical path (hereinafter referred to as the reference theoretical load) (S5005). That is, when the allocation priority of the logical path is increased by the change value, it is checked whether or not the theoretical load of the reference logical path, which is the largest theoretical load, is exceeded. Here, when the comparative theoretical load is smaller than the reference theoretical load (S5005: YES), the path management driver 120 indicates that the allocation priority registered in the logical path management table 400 is a preset allocation priority. It is determined whether it is the upper limit value of the range (S5006). Here, the domain is a value that the allocation priority can take. The range can be arbitrarily set by an operator or the like. The set region is registered in the logical path management table 400. In (S5006), if the content of the allocation priority in the logical path management table 400 is not the upper limit value (S5006: NO), the change value is added to the content of the allocation priority of the logical path in the logical path management table 400. Then, the update counter is incremented (+1) (S5008).
[0031]
On the other hand, in (S5006), when the allocation priority matches the upper limit value of the domain (S5006: YES), the allocation priority of logical paths other than the logical path registered in the logical path management table 400 is determined. The change value is subtracted and the update counter is incremented (+1) (S5007). By doing so, the allocation priority of the logical path is relatively increased as compared with other logical paths. Further, the theoretical load of the logical path increases to the extent that the theoretical load of the reference logical path having the largest theoretical load is not exceeded. Also, by unconditionally increasing the allocation priority of data processing requests to logical paths that have been determined to have a heavy load, load concentration on the logical paths can be suppressed, and load distribution between logical paths can be more appropriately distributed. It becomes possible to plan.
[0032]
In (S5005), when the comparative theoretical load is greater than or equal to the reference theoretical load (S5005: NO), the path management driver 120 sets zero (“0”) in the update counter (S5009).
[0033]
In (S5010), the path management driver 120 repeats the processing from (S5004) for the logical paths for which the above processing is not completed among all the logical paths other than the group of reference logical paths connected to a certain logical unit. Execute (S5010: NO). When the above processing is completed for all logical paths other than the reference logical path (S5010: YES), the path management driver 120 then determines that the allocation priority of any one of the group of logical paths is the allocation priority. It is checked whether or not it has been changed (S5011). If the allocation priority of any logical path has been changed (S5011: YES), the change value is subtracted from the content of the allocation priority of the reference logical path in the logical path management table 400 (S5012).
[0034]
In (S5013), when there is another logical unit (LU) for which the allocation priority is not set (S5013: YES), the path management driver 120 determines a group of logical paths connected to the logical unit. The processing from (S5002) is executed. If there is no other logical unit (S5013: NO), the series of processing ends.
[0035]
In the above processing, the allocation priority can be changed by changing the allocation priority range and the change value. For example, by setting a wide range, or by making the change value a small value (hereinafter referred to as stability setting), the change in the allocation priority with respect to the addition / subtraction of the change value becomes small. By setting the range narrower or setting the change value to a large value (hereinafter referred to as “sensitive setting”), the change in the allocation priority with respect to the addition / subtraction of the change value becomes large.
[0036]
By the way, the allocation priority calculated in the past reflects the load of the logical path in the past. Therefore, as described above, by calculating a new allocation priority based on the newly acquired load and the allocation priority reflecting the previously acquired load, the load on the logical path in the past can be reduced. The allocation priority is calculated in a form that takes into account, and an appropriate allocation priority can be calculated.
[0037]
=== Load distribution ===
The path management driver 120 assigns a data input / output request input from an application program, an operating system, or the like to any one of a plurality of logical paths. This allocation is performed based on the allocation priority set in the logical path management table 400 so that load distribution is realized among the logical paths. Here, a mechanism for sequentially allocating a number (amount) of data input / output requests corresponding to the allocation priority to each logical path as a mechanism for realizing load distribution based on the allocation priority will be described. Note that the mechanism for realizing load balancing is not limited to this. For example, data input / output requests can be preferentially allocated to logical paths having a large allocation priority.
[0038]
FIG. 6 is a flowchart for explaining a process in which the path management driver 120 realizes load distribution by sequentially allocating the number of data input / output requests corresponding to the allocation priority to each logical path.
In this process, the path management driver 120 sequentially receives data input / output requests that are sequentially input, starting from the logical path having the highest allocation priority, by the number corresponding to the allocation priority set for each logical path. Assign to. This process is realized, for example, by adding a process for allocating a number of data input / output requests corresponding to the allocation priority to the logical path to the allocation algorithm based on the round robin method.
[0039]
In FIG. 6, first, when a data input / output request is input (S6001), the path management driver 120 allocates the data input / output request to a logical path set as an allocation destination at that time. Here, the logical path management table 400 manages an allocation destination flag which is information indicating which logical path is set as an allocation destination at that time. “1” is set in the assignment destination flag of the logical path to which the data input / output request is assigned at that time, and “0” is set in the assignment destination flags of the other logical paths. The contents of the assignment destination flag are managed by processing of the path management driver 120 according to the flowchart shown in FIG. The path management driver 120 checks which logical path is set as the allocation destination by referring to the allocation destination flag in the logical path management table 400 (S6002).
[0040]
The path management driver 120 allocates a data input / output request to the logical path set as the allocation destination (S6003). The assignment of the data input / output request specifically refers to registering the data input / output request in the data input / output request processing waiting queue provided for each logical path. Further, information for identifying a path to which a data input / output request is allocated may be associated with the data input / output request and registered in a data input / output request processing queue provided for a plurality of paths. . The path management driver 120 manages a counter (hereinafter referred to as an allocation counter) for counting the number of data input / output requests allocated to the logical path. The path management driver 120 increments (+1) the value of the allocation counter every time one data input / output request is allocated to the logical path (S6004). Here, when allocating the data input / output request to the logical path, the path management driver 120 checks whether the allocation counter has reached the allocation upper limit value (S6005). The allocation upper limit value is a value determined according to the allocation priority. For example, load distribution is performed among four logical paths, the allocation priority range is set to 1-20, and the allocation priority of each logical path is 5, 6, 7, 3. In this case, a value (for example, 5, 6, 7, 3) determined by the ratio of allocation priority is set as the allocation upper limit value of each logical path. If the allocation counter has reached the allocation upper limit value (S6005: YES), the path management driver 120 executes processing for changing the allocation destination flag shown in FIG. 7 (hereinafter referred to as allocation destination flag change processing). (S6006).
[0041]
The process for changing the assignment destination flag will be described with reference to the flowchart shown in FIG. First, the path management driver 120 sets the assignment destination flag to “from among other logical paths belonging to the logical path group to which the logical path whose assignment destination flag is currently set to“ 1 ”belongs. A logical path having an allocation priority equal to or lower than the allocation priority set for the logical path set to “1” is selected (S7001). If such a logical path cannot be selected (S7002: cannot be selected), the path management driver 120 selects a logical path with the highest allocation priority from among the logical paths belonging to the logical path group ( S7003). If there are a plurality of logical paths having the highest allocation priority (S7004: YES), the logical path to which the youngest logical path ID is assigned is selected as the logical path to which the data input / output request is allocated. (S7005). Then, “1” is set to the assignment destination flag of the logical path management table 400 of the logical path (S7006), and “0” is assigned to the assignment destination flag of another logical path belonging to the logical path group to which the logical path belongs. Set. At this time, zero (0) is set in the allocation counter (S7007).
[0042]
If there is only one logical path having the maximum allocation priority in (S7004), that logical path is selected as the logical path to which the data input / output request is allocated (S7008), and the above (S7006) Proceed to the process.
[0043]
In (S7002), when an allocation priority lower than the allocation priority is set (S7002: lower priority), it is checked whether or not there are a plurality of selected logical paths (S7009). If there are a plurality of logical paths selected here (S7009: YES), the logical path with the smallest logical path ID is selected as the logical path to which the data input / output request is assigned (S7010). ), The process proceeds to the above-described (S7006).
[0044]
If an allocation priority equal to the allocation priority is set in (S7002) (S7002: equal priority), the logical path ID is compared with the selected logical path ID (S7011). . If the ID of the logical path is larger than the ID of the selected logical path (S7011: The logical path ID is large), a logical path having an ID larger than the ID of the logical path is selected. The process proceeds to (S7001). In this case, in the process in (S7001), the same ID as that of the selected logical path is not selected. If there is no logical path having an ID larger than that of the logical path, a logical path having an allocation priority lower than the allocation priority of the logical path is selected (S7001). On the other hand, if the ID of the logical path is smaller than the ID of the selected logical path (S7011: The logical path ID is small), the process proceeds to (S7008).
[0045]
=== Adjustment of allocation priority ===
As described above, changing the allocation priority can be adjusted by changing the range of the allocation priority and the value of the change value. For example, if the load on each logical path is relatively stable and the change over time is gradual, set a wide range or set the change value to a small value (stability setting). Thus, the change of the allocation priority with respect to the addition / subtraction of the change value is reduced, so that the load of each logical path is not changed so rapidly with respect to the change of the load. Also, when the load on each logical path is not stable and the change over time is severe, the change range can be added by setting the change range narrower or setting the change value to a large value (sensitive setting). The change of the allocation priority with respect to the subtraction is increased, and the load of each logical path is changed quickly with respect to the change of the load.
[0046]
On the other hand, depending on the use of the information processing system, there may be a time zone in which the load of the logical path is stable and an unstable time zone. In such an information processing system, it is not always necessary to constantly set either a stable setting or a sensitive setting. Alternatively, it may be preferable to dynamically change the allocation priority change mode by changing the change value. Therefore, the path management driver 120 of the present invention has a function of dynamically changing a domain or a change value registered in the logical path management table 400.
[0047]
FIG. 8 is a flowchart for explaining processing for realizing the above functions. First, the path management driver 120 checks whether there is a logical path for which the update counter of each logical path belonging to the logical path group is greater than or equal to a threshold value (S8001). The reason why the value of the update counter of each logical path is compared with the threshold value is to check the update frequency of the allocation priority. If there is a logical path that is equal to or greater than the threshold value, that is, if the allocation priority is frequently changed (S8001: YES), then the logical path group is set to the current stable setting. It is checked whether the setting is sensitive or sensitive (S8002). Here, the setting of the logical path group is managed in the policy item of the logical path management table 400 shown in FIG. If the stability is set (S8002: YES), the policy setting of the logical path group is changed to the sensitive setting by narrowing the range or increasing the change value (S8003). On the other hand, in (S8002), when the policy setting of the logical path group is a sensitive setting, the sensitive setting is maintained as it is. In this case, a more sensitive (sensitive) setting may be obtained by further narrowing the range or increasing the change value.
[0048]
In (S8001), when there is no logical path that is equal to or higher than the threshold value, that is, when the assignment priority is not frequently changed (S8001: NO), the logical path group is currently in stability. It is checked whether the setting is sensitive or sensitive (S8004). If the policy setting of the logical path group is a sensitive setting (S8004: YES), the policy setting of the logical path group is stabilized by expanding the range or decreasing the change value. (S8005). On the other hand, if the policy setting of the logical path group is the stability setting in (S8004) (S8004: NO), the stability setting is maintained as it is. In this case, the range may be further expanded or the change value may be decreased to achieve a more stable (stable) setting.
[0049]
As described above, in the information processing system according to the present embodiment, the load distribution of the logical path group is set according to the change of the allocation priority set according to the state of the load on the logical path, Since the sensitive management is dynamically changed, it is possible to appropriately distribute the load according to the load of the logical path that changes variously.
[0050]
Also, there are various ways of changing the logical path load over time, such as when a constant load occurs constantly, suddenly high load, or when the load gradually changes over time. It is. According to the above mechanism, the amount of increase / decrease in the allocation priority with respect to the change in the load is changed according to the change in time of the load of each logical path. It becomes possible to realize load balancing
When an operator or the like sets a range or change value of an allocation priority, for example, a policy is defined as an option corresponding to the range or change value, or these change modes. A user interface may be provided so that a domain and a change value can be easily set by selecting a policy. Then, by changing the policy, the operator or other user can adjust the allocation priority according to, for example, the past performance or the logical path load that is empirically grasped. It is possible to realize load balancing. In FIG. 4 described above, “stability” which is a policy for setting a large range of allocation priority or a small change value for the logical path group corresponding to LU1 is set for the logical path group corresponding to LU2. “Sensitive”, which is a policy for setting a lower allocation priority range or a larger change value, is “stability” or “sensitive” for logical path groups corresponding to LU3, depending on the change in logical path load. Is set to “smart”, which is a policy that sets the range to the same value as in the case of stability.
[0051]
=== Detailed Configuration Example of Information Processing Apparatus ===
FIG. 9 is a detailed configuration example of the information processing apparatus 100 according to the present embodiment.
The application program 110 accesses the logical unit 220 set in the storage device 200 via the file system 240. The file system 240 is usually provided as one of the functions of the operating system, and refers to software that manages a storage area provided by the storage device 200. That is, an abstract concept of a file is realized in a logical block on a storage area provided by the storage device 200, and a data input / output request for the file is made to correspond to a data input / output request in units of logical block addresses.
[0052]
The path management driver 120 distributes data input / output requests issued from the application program 110 via the file system 240 to the HP ports IF1 (141) to IF4 (144). The distributed data input / output request is transmitted from the IF1 (141) to IF4 (144) to the storage device 200 via the disk driver 160 (161 to 168) and the IF driver 170 (171 to 178). The disk driver 160 abstracts a block mode device as a logical disk. The IF driver 170 converts and restores the data input / output request into a packet of a predetermined communication protocol. As the communication protocol, for example, a fiber channel can be adopted.
[0053]
The data input / output request is transmitted from the information processing apparatus 100 to the storage device 200 through a physical path configured by hardware including the four connection units 301 to 304. On the other hand, as described above, a logical path is associated with the physical path. A logical path is a data path logically configured by software. In FIG. 9, logical paths are indicated by 801 to 808. A logical path is provided for each LU.
[0054]
The information processing apparatus 100 and the storage device 200 can be connected via a SAN (Storage Area Network) 700 as shown in FIG. 9, or can be directly connected as shown in FIG. You can also.
[0055]
The driver management program 121 manages the path management driver 120 together with a driver UI (User Interface) program 122 and a driver API (Application Program Interface) program 123. The driver UI program 122 is a program for providing a user interface when setting the path management driver 120 via the driver management program 121. The driver API program 123 is a program in which instructions and functions that can be used when setting the path management driver 120 using the driver UI program 122 are collected.
[0056]
FIG. 10 shows a part of the configuration different from the information processing apparatus 100 shown in FIG. In the information processing apparatus 100 in FIG. 9, the disk driver 160 receives a data input / output request from the path management driver 120, abstracts a block mode device as a logical disk, and passes it to the IF driver 170. However, some operating systems require the disk driver 160 to be interposed between the file system 240 and the path management driver 120. FIG. 10 shows an example of such a case. The form of FIG. 9 and the form of FIG. 10 are caused by a difference in operating system, and both forms are one embodiment of the present invention.
[0057]
The information processing apparatus and the like according to the present invention have been described above. However, the above-described embodiment is intended to facilitate understanding of the present invention and does not limit the present invention. The present invention can be changed and improved without departing from the gist thereof, and the present invention includes the equivalents thereof.
[0058]
【The invention's effect】
ADVANTAGE OF THE INVENTION According to this invention, the control method of an information processing apparatus, an information processing apparatus, an information processing system, and a program can be provided.
[Brief description of the drawings]
FIG. 1 is a diagram showing an overall configuration of an information processing system according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating a logical path according to an embodiment of the present invention.
FIG. 3 is a flowchart illustrating processing for acquiring a load on a logical path according to an embodiment of the present invention.
FIG. 4 is a diagram showing a logical path management table 400 according to an embodiment of the present invention.
FIG. 5 is a flowchart illustrating processing for setting an allocation priority of a logical path management table 400 according to an embodiment of the present invention.
FIG. 6 is a flowchart illustrating a process for realizing load distribution by sequentially allocating a number of data input / output requests according to an allocation priority to each logical path according to an embodiment of the present invention.
FIG. 7 is a flowchart illustrating processing for changing an assignment destination flag according to an embodiment of the present invention.
FIG. 8 is a flowchart illustrating processing for realizing a function of dynamically changing a domain or a change value registered in a logical path management table 400 according to an embodiment of the present invention.
FIG. 9 is a diagram illustrating a detailed configuration example of an information processing apparatus 100 according to an embodiment of the present invention.
FIG. 10 is a diagram illustrating a detailed configuration example of an information processing apparatus 100 according to an embodiment of the present invention.
[Explanation of symbols]
100 Information processing apparatus
110 Application program
120 path management driver
131 to 138 Data input / output queues assigned to logical paths
141-148 IF (Inter Face)
200 storage device
210 Disk controller
220 logical units
400 Logical path management table

Claims (15)

In a control method of an information processing apparatus for transmitting a data input / output request to a storage device through a plurality of logical paths set between the storage devices connected to be communicable,
Obtain the load of the logical path;
Determining the logical path to which the data input / output request is assigned according to the load;
A method for controlling an information processing apparatus. The control method of the information processing apparatus according to claim 1,
An allocation priority that is a priority when allocating the data input / output request is calculated for the logical path based on the load,
Allocating the data input / output request to the logical path in an amount corresponding to the allocation priority calculated for the logical path;
A method for controlling an information processing apparatus. In the control method of the information processing apparatus according to claim 2,
Updating the allocation priority based on a value calculated based on the acquired load of the logical path and the allocation priority calculated in the past corresponding to the logical path;
A method for controlling an information processing apparatus. The information processing apparatus according to claim 2,
The calculation of the allocation priority is as follows:
The allocation priority is increased / decreased according to the load, and the allocation priority increase / decrease amount with respect to the load change is changed according to a temporal change mode of the load of each logical path,
A method for controlling an information processing apparatus. In an information processing apparatus that transmits a data input / output request to the storage device through a plurality of logical paths set between the storage devices that are communicably connected,
Means for obtaining a load of the logical path;
Means for determining the logical path to which the data input / output request is assigned according to the load;
An information processing apparatus comprising: The information processing apparatus according to claim 5,
Means for calculating an allocation priority, which is a priority when allocating the data input / output request, for the logical path based on the load;
Means for allocating to the logical path an amount of the data input / output request corresponding to the allocation priority calculated for the logical path;
An information processing apparatus comprising: The information processing apparatus according to claim 6,
Means for updating the allocation priority based on a value calculated based on the obtained load of the logical path and the allocation priority calculated in the past corresponding to the logical path;
An information processing apparatus characterized by the above. The information processing apparatus according to claim 6,
The calculating means is means for increasing / decreasing the allocation priority according to the load, and comprises an user interface for designating an increase / decrease amount of the allocation priority with respect to a change in the load. . The information processing apparatus according to claim 6,
The calculating means increases / decreases the allocation priority according to the load, and changes the allocation priority increase / decrease amount with respect to the load change according to a temporal change mode of the load of each logical path. Being
An information processing apparatus characterized by the above. The information processing apparatus according to any one of claims 5 to 9,
The information processing apparatus, wherein the load of each logical path is calculated based on a response time of a data input / output request transmitted through each logical path. The information processing apparatus according to any one of claims 5 to 9,
The logical path is a communication path defined by a combination of one or more software or one or more hardware connecting the information processing apparatus and the storage device. . Including an information processing device and a storage device that are communicably connected;
Means for transmitting a data input / output request to the storage device through a plurality of logical paths set between the information processing device and the storage device; and acquiring the load of each logical path to obtain the data input / output request. And a means for determining the logical path to be assigned according to the load. To an information processing device that transmits a data input / output request to the storage device through a plurality of logical paths set between the storage devices that are communicably connected,
A function of acquiring a load of the logical path;
A function of determining the logical path to which the data input / output request is assigned according to the load;
Program to realize. In a control method of an information processing apparatus for transmitting a data input / output request to a storage device through a plurality of logical paths set between the storage devices connected to be communicable,
The time from when the data input / output request is transmitted to the logical path until the response to the data input / output request is received is obtained as a response time,
An allocation priority that is a priority when allocating the data input / output request is calculated for the logical path based on the response time;
A theoretical load that is a value calculated from the response time and the assigned priority is calculated for the logical path,
The logical path having the maximum theoretical load among the calculated theoretical loads of the logical path is selected as a reference logical path,
A comparison theoretical load that is a value calculated from the response time and a value obtained by adding a predetermined change value to the allocation priority is calculated for the logical path;
When the logical load for the reference logical path is larger than the comparative theoretical load, and the allocation priority of the logical path is the upper limit value of the allocation priority range, the logical path other than the logical path The change value is subtracted from the allocation priority, and when the logical path allocation priority is not the upper limit value of the allocation priority range, the change value is added to the logical path allocation priority,
When the allocation priority of any of the logical paths is changed, the change value is subtracted from the allocation priority of the reference logical path,
Allocating the data input / output request in an amount corresponding to the allocation priority to the logical path;
A method for controlling an information processing apparatus. In a control method of an information processing apparatus for transmitting a data input / output request to a storage device through a plurality of logical paths set between the storage devices connected to be communicable,
Obtain the load of the logical path;
An allocation priority that is a priority when allocating the data input / output request is calculated for the logical path based on the load,
Updating the allocation priority based on a value calculated based on the acquired load of the logical path and the allocation priority calculated in the past corresponding to the logical path;
Check the update frequency of the allocation priority of the logical path,
When the allocation priority of the logical path is frequently changed, the allocation priority range is changed to be narrow or the allocation priority change value is changed to a large value,
When the allocation priority of the logical path is not frequently changed, the domain is changed so that the domain becomes wider or the change value becomes a small value.
Allocating the data input / output request to the logical path in an amount corresponding to the allocation priority calculated for the logical path;
A method for controlling an information processing apparatus.

Images