JP4686564B2 - Load balancing system - Google Patents

Description

  The present invention relates to a load distribution system applied to a system in which a plurality of access paths exist between a host computer and a storage device.

  Conventionally, in a system in which a single storage device is provided for a host computer and multiple access paths exist between them, input / output is distributed over multiple access paths to efficiently use the access paths. There are some that do load balancing.

  However, according to such a system using load balancing, a load balancing algorithm that uses a round robin method that sends generated I / O requests to multiple access paths in a round robin manner is, for example, continuous If the so-called sequential read, in which data is read sequentially from the designated area, is load-balanced, the system performance may deteriorate. There are two causes, and the first is a storage device, for example, a disk array device having a plurality of controllers and a single LU (Logical Unit). When issued to each controller distributed to the path, the controller that receives the distributed request is alternately input a read request, and when viewed from each controller side, the data area of the read request becomes discontinuous, In some cases, the accuracy of determining that the read is sequential decreases, and the probability that prefetching fails will increase. Second, each controller may read multiple times to the same address of a single LU, increasing the load on the disk.

Therefore, for example, as disclosed in Patent Document 1, a data storage device is provided for a host computer, random access uses a plurality of access paths to distribute the load, and a single read request for sequential read requests. Some use a path.
JP 2003-99384 A

  However, in Patent Document 1, since a single access path is used for sequential read and load distribution cannot be performed, read / write requests are concentrated on the controller connected to the single access path, causing an increase in load and The performance of the system may be degraded, for example, there is no room in the bandwidth used for one access path.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a load distribution system that enables good load distribution to various leads including sequential reads.

The load distribution system according to the present invention is:
A host computer,
A first storage device having a plurality of controllers and a storage unit body whose reading and writing are controlled by the controllers;
A second storage device including a plurality of controllers and a storage unit main body storing duplicate data of data written to the first storage device, the reading and writing of which are controlled by the controller;
A plurality of access paths connected between the host computer and the controllers of the first and second storage devices;
Comprising
In response to a sequential read request, the host computer selects the access path for each of the controllers of the first and second storage devices, and distributes the requests to the first and second storage devices. Issue ,
Each of the first and second storage devices has a read table in which a reading range is recorded in each controller;
In response to the sequential read request, the host computer selects the access path of the controller in which the read table exists according to the comparison result between the sequential read request range and the read range recorded in the read table. The requests are distributed to the first and second storage devices .

  According to the present invention, it is possible to provide a load distribution system that enables good load distribution for various leads including sequential reads.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

(First embodiment)
FIG. 1 shows an overall block diagram of a load distribution system according to a first embodiment of the present invention. In FIG. 1, reference numeral 100 denotes a host computer, and a master disk array device 200 as a first storage device and a sub disk array device 300 as a second storage device are connected to the host computer 100, respectively. The master disk array device 200 includes two controllers 210 and 220 and an LU (Logical Unit) 230 as a storage unit main body, and the sub disk array device 300 includes two controllers 310 and 320 and an LU (Logical Unit). 330. The controllers 210 and 220 of the master disk array apparatus 200 are connected to the host computer 100 via FC (Fiber Channel) cables 401 and 402 as access paths, respectively, and write to and read from the LU 230 according to the request of the host computer 100. To control. Similarly, the controllers 310 and 320 of the sub disk array apparatus 300 are also connected to the host computer 100 via the FC cables 403 and 404, respectively, and control writing and reading to the LU 330 according to instructions from the host computer 100. In this case, the LU 330 has a copy of the data written in the LU 230.

  The controller 210 of the master disk array device 200 and the controller 310 of the sub disk array device 300 are connected by an FC cable 411. Similarly, the controller 220 of the master disk array device 200 and the controller 320 of the sub disk array device 300 are connected to each other. Are connected by an FC cable 412. These FC cables 411 and 412 are used for communication between the master disk array device 200 and the sub disk array device 300.

  Note that the LU 230 of the master disk array device 200 and the LU 330 of the sub disk array device 300 have the same capacity in a one-to-one pair, and these LU 230 and LU 330 are mirrored synchronously.

  FIG. 2 shows a partial block diagram of the host computer 100.

  In this case, the host computer 100 is roughly composed of an application 101, a file system 102, a load distribution driver 110, and physical devices (device files) 131 to 134 as software configurations. The application 101 reads / writes data from / to the master disk array device 200 and the sub disk array device 300 via the file system 102 and below.

  The file system 102 converts a read / write request for a file into a block-level read / write request. Here, the block level read / write request includes information such as read / write attributes, read / write sector addresses, and the number of read / write sectors. Further, in a general operating system, the file system 102 performs prefetching when it detects that a plurality of read areas for a read request are continuous, and operates to hold the prefetched data in the read cache. Therefore, the file system 102 here converts the read area of the discontinuous read request issued by the application 101 into a sequential read request for reading large continuous data.

  The load distribution driver 110 divides and distributes data read / write requests to physically connected access paths (corresponding to the FC cables 401 to 404 described above). In the physical devices 131 to 134, the LU 230 of the master disk array device 200 and the LU 330 of the sub disk array device 300 are recognized as device files on the host computer 100. The above-described FC cables 401 to 404 are connected to the physical devices 131 to 134, respectively.

  The load distribution driver 110 includes a distribution control unit 111, an I / O type determination unit 112, and a path information table 113. The distribution control unit 111 controls load distribution and will be described in detail later.

  The I / O type determination unit 112 determines the type of the read / write request. In this case, the I / O type determination unit 112 performs a write request (including both random and sequential), random read (random read request), sequential read (sequential read) based on the read / write request passed from the distribution control unit 111. One of the three patterns (read request) is determined. Here, the sequential read is a method of reading data sequentially from a continuous area, and the random read is a method of reading data from an arbitrary position regardless of the order.

  Whether the request is a read request or a write request is determined from the read / write attribute included in the request. Whether the read is sequential read or random read is determined as sequential read when the number of read sectors included in the read request information is a certain number or more (for example, 1024 sectors), otherwise random read. Is determined.

The path information table 113 holds access path information for all the disk array devices connected to the host computer 100, that is, the master disk array device 200 and the sub disk array device 300.

  Table 1 shows an example of the path information table 113, and has items of “path number”, “device type”, “controller number”, and “use flag”. In the “path number”, serial numbers 0, 1, 2, and 3 assigned for each access path, that is, for each of the FC cables 401 to 404 described above are written. The “device type” corresponds to the type of the disk array device corresponding to the path numbers 0, 1, 2, and 3, and here, the path numbers 0 and 1 (path numbers assigned to the FC cables 401 and 402). Thus, “master” representing the master disk array device 200 and “sub” representing the sub disk array device 300 are written in correspondence with the path numbers 2 and 3 (pass numbers assigned to the FC cables 403 and 404). The “controller number” is associated with the path numbers 0 and 1, “0” and “1” respectively representing the controllers 210 and 220 of the master disk array apparatus 200, and the sub disk array apparatus 300 is associated with the path numbers 2 and 3. “0” and “1” respectively representing the controllers 310 and 320 are written. Furthermore, “ON” or “OFF” as shown in the figure indicating the usage status of the controllers 210, 220, 310, 320 corresponding to the pass numbers 0, 1, 2, 3 is written in the “usage flag”. Yes. Such a path information table 113 is based on information from all the disk array devices connected to the host computer 100, that is, the master disk array device 200 and the sub disk array device 300 when the host computer 100 is started. Generated.

  Next, the operation of the embodiment configured as described above will be described.

  First, data reading / writing by the host computer 100 will be briefly described.

  In the case of data writing, for example, when the host computer 100 requests the controller 210 of the master disk array device 200 to write data to the LU 230, the controller 210 writes data to the LU 230. When data writing to the LU 230 is completed, the controller 310 of the sub disk array device 300 is requested to write data to the LU 330 via the FC cable 411. When the data writing to the LU 330 is completed, the controller 310 notifies the controller 210 of the completion of writing via the FC cable 411. Upon receiving this notification, the controller 210 notifies the host computer 100 of the completion of writing and writes the data. To complete.

  Further, the host computer 100 performs a data write request to the LU 230 to the controller 220 of the master disk array device 200 in the same manner as described above.

  As described above, the data write request by the host computer 100 is accepted only by the controllers 210 and 220 of the master disk array device 200, and the acceptance by the controllers 310 and 320 of the sub disk array device 300 is rejected.

  Next, in the case of data reading, for example, when the host computer 100 requests the controller 210 of the master disk array device 200 to read data, the controller 210 reads data from the LU 230, and when this reading is completed, The host computer 100 is immediately notified of the completion of data reading.

  Such a data read request can be accepted by all the controllers 210, 220, 310, 320 of the master disk array device 200 and the sub disk array device 300. Therefore, in this state, communication between the master disk array device 200 and the sub disk array device 300 via the FC cables 411 and 412 is not performed.

  In this way, the host computer 100 can request the master disk array device 200 to read / write data, and can request only the sub disk array device 300 to read data.

  Next, load distribution for various requests will be described.

  In this case, the load distribution driver 110 is mounted on the host computer 100, and the sequential read and random read are distributed and issued to the master disk array device 200 and the sub disk array device 300, thereby reducing the load.

  Now, when the application 101 of the host computer 100 issues a data read / write request to the master disk array device 200 and the sub disk array device 300, the file system 102 receives this data read / write request.

  The file system 102 converts the read / write request into a block level read / write request including information such as a read / write attribute, a read / write sector address, and the number of read / write sectors. The converted block level read / write request is passed to the distribution control unit 111. The distribution control unit 111 passes a read / write request to the I / O type determination unit 112, and the I / O type determination unit 112 uses any one of the three patterns of write request (including both random and sequential), random read, and sequential read. Determine whether. In this case, whether the request is a read request or a write request is determined from the read / write attribute included in the request. Whether the read request is a sequential read or a random read is determined as a sequential read when the number of read sectors included in the read request information is a certain number (for example, 1024 sectors). Otherwise, it is determined as a random read.

  The determination result in the I / O type determination unit 112 is passed to the distribution control unit 111, and the distribution control unit 111 determines an access path to be distributed according to the above three patterns. For example, when the I / O type determination unit 112 determines that the read / write request is neither a sequential read nor a random read, that is, a write request, only the controllers 210 and 220 of the master disk array device 200 can accept them as described above. Therefore, the distribution control unit 111 refers to the path information table 113 in Table 1, and controller numbers “0” and “1” of the master disk array device 200 corresponding to “0” and “1” of “path number”. The controllers 210 and 220 are selected, and the physical devices 131 and 132 corresponding to the controllers 210 and 220 are distributed to the FC cables 401 and 402 (access paths) to issue requests. As this distributed algorithm, a round robin method of issuing alternately, a method of issuing to a path with few requests in the queue, and the like can be considered, and one of these distributed algorithms is selected and used.

  Next, when the I / O type determination unit 112 determines that the read / write request is a random read, as described above, all the controllers 210, 220, 310 of the master disk array device 200 and the sub disk array device 300 are used. 320, the distribution control unit 111 refers to the path information table 113 in Table 1, and the master disk array device 200 and the sub disk array device corresponding to “0” to “3” of “path number”. All of the controllers 210, 220, 310, and 320 with controller numbers “0” and “1” of 300 are selected, and FC cables are connected from the physical devices 131, 132, 133, and 134 corresponding to these controllers 210, 220, 310, and 320. Issue requests distributed to 401-404 (access path) That. In this case, as the distributed algorithm, either the round robin method or the method of issuing to a path with few requests in the queue is selected and used.

  Next, when the I / O type determination unit 112 determines that the read / write request is sequential read, also in this case, as described above, all the controllers 210 and 220 of the master disk array device 200 and the sub disk array device 300 are also used. 310 and 320, the distribution control unit 111 refers to the path information table 113 in Table 1 and executes the flowchart shown in FIG. 3 to determine the access path of the distribution destination. In this case, in FIG. 3, first, in step 301, the controller number X is set to “0”, and in step 302, the path information table 113 is referred to and the controller number is “0”. 2 is referred to. Next, in step 303, it is determined as no due to the presence of the controller X (= “0”). In this case, since the use flags of the rows with pass numbers 0 and 2 are both ON, it is determined as yes, and the process proceeds to step 305, (“X + 1” → X (= “1”), and returns to step 302. Then, the path information table 113 is referred again, and this time, the line with the controller number “1”, here, the line with the path numbers 1 and 3 is referred to, and at step 303, the controller X (= “1”). ) Is determined to be no, and the process proceeds to step 304 to determine whether or not the use flag of the line with the controller number “1” is ON. Therefore, it is determined as no, and the process proceeds to Step 306, where the controller 220 of the master disk array device 200 with the control number “1” and the sub disk array device 300 are controlled. The use flag of the controller 320 with the control number “1” is turned on, and the turned-on path numbers 1 and 3 are output as a result, whereby the FC cables 402 and the physical devices 132 and 134 corresponding to the controllers 220 and 320 are output. A request is issued in a distributed manner in 404 (access path), and the distribution algorithm in this case selects the round robin method, which corresponds to “path numbers” 0, 1, 2, and 3 in the path information table 113. If the use flags are all “OFF”, for example, in step 302, the path information table 113 is referred to and the line “controller number” is “0”, here the lines with path numbers 0 and 2 are referred to. When the process proceeds to step 304 through step 303 and it is determined whether the use flag of the line with the controller number “0” is ON, the pass number Since the use flags of the second row are both OFF, the process immediately proceeds to step 306, where the controller 210 with the control number “0” of the master disk array device 200 and the controller 310 with the control number “0” of the sub disk array device 300 The use flag is set to ON, and the path numbers that are set to ON are output as a result of 0. As a result, the physical devices 131 and 133 corresponding to the controllers 210 and 310 are distributed to the FC cables 401 and 403 (access paths). In this case as well, the distributed algorithm selects the round robin method.

  In the flowchart of FIG. 3, if it is determined in step 303 that the controller X does not exist, the process proceeds to step 307, where one pass number is randomly selected from the pass numbers and a request is issued.

  Finally, when the sequential read request is completed, the flag recorded in the “use flag” in the path information table 113 is cleared to the original state.

  Therefore, in this case, the two master disk array devices 200 and the sub disk array device 300 are connected as storage devices to the host computer 100 in which the load distribution driver 110 is mounted, and the master disk array device 200 and the sub disk array device 200 are connected. Various requests are distributed and issued to the disk array device 300. That is, in the case of a data write request, the access paths of the controllers 210 and 220 of the master disk array device 200 are selected and distributed and issued. In the case of a random read request, the master disk array device 200 and the sub disk array device are issued. Access paths of all the controllers 210, 220, 310, and 320 of 300 are selected, issued in a distributed manner, and in the case of a sequential read request, one each from the master disk array device 200 and the sub disk array device 300. The access path is selected, distributed, and issued so that a good load distribution can be achieved for all the leads. In particular, in the case of a sequential read request, by referring to the path information table 113 and executing the flowchart shown in FIG. 3, the controllers 210 and 310 with the controller number “0” of the master disk array device 200 and the sub disk array device 300, respectively. Or a combination pattern of the controllers 220 and 320 having the controller number “1” in each of the master disk array device 200 and the sub disk array device 300, and each access path is selected for each of these patterns. Since sequential read requests can be issued in a distributed manner, sequential read performance can be improved. Further, in this sequential read, the request issue destination can be distributed to a plurality of different disk array devices (the master disk array device 200 and the sub disk array device 300). The controller 210 (220) of the master disk array device 200, The controller 310 (320) of the sub disk array apparatus 300 does not read data from the same area of the same LU multiple times, and an increase in the load on the LU can be prevented. Further, in the sequential read, when the master disk array device 200 and the sub disk array device 300 support the read ahead with respect to the sequential read, if the sequential read is divided by the round robin method, the data area of the read request is not available. Since it becomes continuous, the detection accuracy of sequential reads may be lowered. However, in the master disk array device 200 and the sub disk array device 300, discontinuous and unidirectional read requests can be detected as sequential reads. There is no problem. In this case, if the sequential read is divided by the round robin method, the divided requests are discontinuous, but the sector address of the received request always increases, so it is determined as sequential access by referring to the past request history. It's easy to do.

  In the case of a random read request, the access paths of all the controllers 210, 220, 310, and 320 of the master disk array device 200 and the sub disk array device 300 can be selected and distributed and issued. Random reading performance can also be improved.

  In the first embodiment, the host computer 100 and the master disk array device 200 and the sub disk array device 300 are connected by FC cables 401 to 404, but instead, via an FC switch. May be connected. In FIG. 1, for the sake of convenience, the master disk array device 200 has two controllers 210 and 220, and the sub disk array device 300 also has two controllers 310 and 320. The sub-disk array device 300 may have three or more controllers. The master disk array device 200 and the sub disk array device 300 show a case where one set of LU 230 and LU 330 having the same capacity exists, but the LU in the master disk array device 200 and the LU in the sub disk array device 300 are shown. There may be a plurality of LU pairs as long as the conditions of the same capacity are satisfied and the condition is that they are one-to-one pairs. If there is a plurality of LUs, a load information can be distributed for each LU by providing a path information table for each LU. Further, FIG. 1 shows the case of one host computer 100 for the sake of simplicity, but a plurality of units can be connected by using an FC switch. When a plurality of host computers 100 are connected, since different host computers generally use different LUs, the load can be distributed appropriately.

(Second Embodiment)
Next, a second embodiment of the present invention will be described.

  FIG. 4 is an overall block diagram of the load distribution system according to the second exemplary embodiment of the present invention. The same parts as those in FIG. 1 and FIG.

In this case, the read table 211, 221 is connected to each controller 210, 220 in the master disk array device 200, and the read table 311, 321 is connected to each controller 310, 320 in the sub disk array device 300. ing. As shown in Table 2, the read table 211 is a table having fields of “first sector address” and “number of sectors” indicating the read range, and the initial value is 0 (indicated by φ in the flowchart shown in FIG. 5). ). The same applies to the other lead tables 221, 311, 321. Further, an algorithm shown in FIG. 5 for determining an access path for load distribution is added to the distribution control unit 111.

  Others are the same as FIG.1 and FIG.2.

  In this case, if the request is determined to be a write request or random read by the I / O type determination unit 112, the distribution control unit 111 selects a distribution destination access path as described in the first embodiment. , Issue requests in a distributed manner.

  Next, when the I / O type determination unit 112 determines that the read / write request is sequential read, the distribution control unit 111 determines an access path for load distribution by executing the flowchart shown in FIG.

  In this case, the access path is selected in the following four patterns (1) to (4). (1) When there are empty read tables in the master disk array device 200 and the sub disk array device 300, respectively, in this case, the controller 210 of the master disk array device 200 with the controller number (i) = 0, the sub disk array device A controller 310 having a controller number (j) = 0 of 300 is used, and the read range Tm (i) of the read table 211 of the controller 210 and the read range Ts (j) of the read table 311 of the controller 310 are the initial values φ. .

  In this state, in step 501, the distribution control unit 111 reads the sequential read range R, the read range Tm (i) recorded in the read table 211 of the controller 210 of the master disk array device 200, and the sub disk array device 300. The reading range Ts (j) recorded in the read table 311 of the controller 310 is sequentially read. Here, the read range R is obtained from the read sector address and the number of sectors passed from the file system 102.

  In step 502, the sequential read range R and the read range Tm (i) of the read table 211 are compared, and the sequential read range R and the read range Ts (j) of the read table 311 are compared. In this case, the reading ranges Tm (i) and Ts (j) are both initial values φ, and the reading ranges Tm (i) and Ts (j) of these initial values φ do not include the reading range R. No is determined, and the process proceeds to step 503. Also in this step 503, the sequential read reading range R and the reading range Tm (i) of the read table 211 are compared, and the sequential reading reading range R and the reading range Ts (j) of the read table 311 are respectively compared. In this case, the read ranges Tm (i) and Ts (j) are both initial values φ, and the read ranges Tm (i) and Ts (j) of these initial values φ may be included in the sequential read read range R. Since no, it is judged as no and the process proceeds to step 504.

  In step 504, since the reading ranges Tm (i) and Ts (j) are both the initial value φ and empty lead tables 211 and 311 exist, it is determined yes and the process proceeds to step 505. In step 505, the sequential read read range R is recorded in the read range Tm (i) of the read table 221 and the read range Ts (j) of the read table 311, and the controllers 210 and 310 in which the read tables 221 and 311 exist are recorded. Select an access path.

  (2) This is a case where a read table having a read range including a sequential read read range R exists in the master disk array device 200 and the sub disk array device 300, respectively. Also in this case, the controller 210 with the controller number (i) = 0 of the master disk array device 200 and the controller 310 with the controller number (j) = 0 of the sub disk array device 300 are used. From this state, in step 501, the distribution control unit 111 reads the sequential read read range R, the read range Tm (i) of the read table 211 of the controller 210 of the master disk array device 200, and the controller of the sub disk array device 300. The reading range Ts (j) of the 310 read table 311 is read in order.

  In step 502, the sequential read range R and the read range Tm (i) of the read table 211 are compared, and the sequential read range R and the read range Ts (j) of the read table 311 are compared. In this case, since the read ranges Tm (i) and Ts (j) include the read range R of the sequential read, it is determined to be yes, and the process proceeds to step 506, and the read having these read ranges Tm (i) and Ts (j). The access paths of the controllers 210 and 310 where the tables 211 and 311 exist are selected.

  (3) Contrary to the above pattern (2), the sequential read read range R includes the read table read ranges of the master disk array device 200 and the sub disk array device 300. Also in this case, the controller 210 with the controller number (i) = 0 of the master disk array device 200 and the controller 310 with the controller number (j) = 0 of the sub disk array device 300 are used. From this state, in step 501, the distribution control unit 111 reads the sequential read read range R, the read range Tm (i) recorded in the read table 211 of the controller 210 of the master disk array device 200, and the sub disk array device. The reading range Ts (j) recorded in the read table 311 of the controller 310 of 300 is sequentially read. In step 502, the sequential read range R and the read range Tm (i) of the read table 211 are compared, and the sequential read range R and the read range Ts (j) of the read table 311 are compared. In this case, since the read ranges Tm (i) and Ts (j) do not include the sequential read read range R, it is determined to be no, and the process proceeds to step 503 to read the sequential read read range R and the read range 211 of the read table 211. (I) The sequential read range R and the read range 311 of the read table 311 are respectively compared. In this case, since the reading ranges Tm (i) and Ts (j) are both included in the sequential reading range R, it is determined to be yes, and the process proceeds to step 505, and the sequential reading range R is read as the reading range Tm (i). ), Ts (j) is overwritten, and the access paths of the controllers 210 and 310 in which the read tables 211 and 311 having these read ranges Tm (i) and Ts (j) exist are selected.

  (4) If none of the above conditions (1) to (3) is met, the process proceeds to step 507 via steps 501 to 504, and one access path is selected at random.

  The controller 210 of the master disk array device 200 and the controller 310 of the sub disk array device 300 clear the information in the read tables 211 and 311 when the data read request within the range recorded in the respective read tables 211 and 311 is completed.

  In the above description, the controller 210 is used as the controller (i) and the controller 310 is used as the controller (j). However, the same applies to the case where the controller 220 is used as the controller (i) and the controller 320 is used as the controller (j). it can.

  Therefore, even if it does in this way, the effect similar to 1st Embodiment can be acquired. Further, read tables 211 and 221 are provided in the controllers 210 and 220 of the master disk array apparatus 200, and read tables 311 and 321 are provided in the controllers 310 and 320 of the sub disk array apparatus 300, respectively. , 321, the write ranges Tm (i) and Ts (j) for determining the access path are recorded. As a result, the master disk array device 200 and the sub disk array device 300 can clearly know the sequential read range, so even if the sequential read is divided into discontinuous requests by load distribution, the sequential read determination accuracy is high. The problem of prefetch failure due to the decrease can be solved. Further, when there are a plurality of host computers, the plurality of host computers can grasp information for determining the same access path by referring to the contents of the respective read tables 211, 221, 311, 321. When performing a plurality of sequential reads on a region, a plurality of host computers can use the same controller 210, 220, 310, 320, and a read cache (not shown) of the controllers 210, 220, 310, 320 is provided. Can also be used efficiently.

  If a failure such as a cable disconnection between the host computer 100 and the master disk array device 200 or between the host computer 100 and the sub disk array device 300 occurs during reading, the respective read tables 211, 221, Information of 311 and 321 may not be cleared permanently. For this reason, the lead tables 211, 221, 311, and 321 are provided with a timeout. That is, the controllers 210, 220, 310, and 320 having the read tables 211, 221, 311, and 321 each have a timer, and this timer always counts up the numerical value of the register and requests from the host computer 100. When received, the register is cleared to zero. When the request from the host computer 100 stops and the count value of the register overflows, the read table is cleared to zero. It is desirable to set the timeout time to about 1 minute. In the second embodiment, for the sake of simplicity, the master disk array device 200 and the sub disk array device 300 are each provided with one LU 230 and 330, but each LU is provided with a read table. It is also possible to use a plurality of LUs.

  In addition, this invention is not limited to the said embodiment, In the implementation stage, it can change variously in the range which does not change the summary.

  Furthermore, the above embodiments include inventions at various stages, and various inventions can be extracted by appropriately combining a plurality of disclosed constituent elements. For example, even if some constituent requirements are deleted from all the constituent requirements shown in the embodiment, the problem described in the column of the problem to be solved by the invention can be solved, and is described in the column of the effect of the invention. If the above effect is obtained, a configuration from which this configuration requirement is deleted can be extracted as an invention.

1 is an overall block diagram of a load distribution system according to a first embodiment of the present invention. The partial block diagram of the host computer used for 1st Embodiment. 6 is a flowchart for explaining access path distribution in the case of sequential read according to the first embodiment; The whole block diagram of the load distribution system concerning the 2nd Embodiment of this invention. 12 is a flowchart for explaining access path distribution in the case of sequential read according to the second embodiment;

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 ... Host computer, 101 ... Application 102 ... File system, 110 ... Load distribution driver 111 ... Distribution control part, 112 ... I / O type determination part 113 ... Path information table, 131-134 ... Physical device 200 ... Master disk array apparatus 210.220, 310.320 ... Controller 211.221, 311.321 ... Read table 230.330 ... LU
300 ... Sub disk array device 401-404, 411, 412 ... FC cable

Claims (4)

A host computer,
A first storage device having a plurality of controllers and a storage unit body whose reading and writing are controlled by the controllers;
A second storage device including a plurality of controllers and a storage unit main body storing duplicate data of data written to the first storage device, the reading and writing of which are controlled by the controller;
A plurality of access paths connected between the host computer and the controllers of the first and second storage devices;
Comprising
In response to a sequential read request, the host computer selects the access path for each of the controllers of the first and second storage devices, and distributes the requests to the first and second storage devices. Issue ,
Each of the first and second storage devices has a read table in which a reading range is recorded in each controller;
In response to the sequential read request, the host computer selects the access path of the controller in which the read table exists according to the comparison result between the sequential read request range and the read range recorded in the read table. A load distribution system for distributing and issuing requests to the first and second storage devices . A host computer,
A first storage device having a plurality of controllers and a storage unit body whose reading and writing are controlled by the controllers;
A second storage device including a plurality of controllers and a storage unit main body storing duplicate data of data written to the first storage device, the reading and writing of which are controlled by the controller;
A plurality of access paths connected between the host computer and the controllers of the first and second storage devices;
Comprising
In response to a sequential read request, the host computer selects the access path for each of the controllers of the first and second storage devices, and distributes the requests to the first and second storage devices. Issue ,
The host computer has a path information table for holding information on access paths to the first and second storage devices, and refers to the path information table to determine the access path of the controller in response to a sequential read request. A load distribution system that selects and issues requests to the first and second storage devices in a distributed manner . The path information table includes at least a “device type” that represents each of the first and second storage devices, a “controller number” that represents a controller provided for each of the first and second storage devices, and a usage for each controller. It has a “use flag” that represents the situation,
The host computer sets a controller number in the path information table in response to a sequential read request, determines the state of a use flag corresponding to the set controller number, and determines the access path of the controller based on the determination result. 3. The load distribution system according to claim 2 , wherein a request is issued in a distributed manner to the first and second storage devices. The load distribution system according to any one of claims 1 to 3 , wherein the host computer issues a request distributed to each of the first and second storage devices in response to a random read request. .

Images