JP2006268708A - Computer system having redundancy connecting structure - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve the reading performance in input and output, in a structure having a host computer connected to an auxiliary storage device through duplexing connection having double route of an active system and a waiting system, by utilizing the bands of both the systems. <P>SOLUTION: The host computer efficiently executes look-ahead from a storage device 106, by assigning look-ahead to an input and output part (operation system) 102 or an input and output part (waiting system) 103, referring to look-ahead entry and considering the order of the request read. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a computer system having a redundant connection configuration, and more particularly to a computer system having a redundant connection configuration that realizes high speed by a prefetching operation.

  In a configuration in which the path is duplicated between the active system and standby system with redundant connections, if there is no failure in either system, the standby system is basically used without operation except for continuity confirmation. It has not been done, and it cannot be said that the entire system is effectively utilizing the bandwidth. Also, in a configuration that does not divide into active and standby systems due to dynamic load balancing, the bandwidth can be used effectively, but there is also a management effort, etc., so load balancing may not be achieved. How to distribute the processing was a big issue.

  In addition, since the continuity of a file on a storage device cannot basically be known except by a file system, it is difficult to implement prefetching independently at a lower hierarchy than the file system.

  In a configuration where the host computer and auxiliary storage device are redundantly connected and the operating system and standby system are duplicated, the normal standby system is used for actual data transfer except for the path continuity check such as heartbeat. Because there is nothing, basically it is often not working. In other words, it was not possible to use the bandwidth reserved for both the active and standby systems.

In addition, there is a method to achieve dynamic load distribution by using redundant paths at the same time without clearly dividing the active and standby systems, but complicated control is required, such as managing which path is allocated. It was not easy to implement.
Furthermore, files that are logical units are not continuously arranged on an actual storage device, and it is necessary to understand the file system by a storage device or device driver itself that should not handle the file system originally. The look-ahead itself was difficult to realize.

  A computer system having a redundant connection configuration according to the present invention includes a host computer, an operational input / output unit, and a standby input / output unit, and the computer has means for performing a prefetch operation on the standby input / output.

  By performing the pre-reading operation, it is possible to conceal the delay in reading when continuous reading is performed on the same file, which can contribute to the improvement of the performance as a whole. Further, by using the standby system, it is possible to execute prefetching without increasing the load on the operational system.

  In addition, since it is not necessary to manage requests for each system, it is possible to simplify the effort for load distribution.

  Next, the best mode for carrying out the present invention will be described in detail with reference to the drawings.

  First, the gist of the present invention will be described.

  In the present invention, when the requested reading is performed from the host computer, the requested block is not only read by the operation route, but is also read by using the standby route at the same time. This prefetching operation is performed along with the operation as a file system managed by the operating system. Only when it recognizes that both the active and standby paths exist, the prefetching operation using the standby path is performed. . However, the pre-reading operation is not performed in a state where the one-system operation in which only one of the active system and the standby system can be used due to a failure or the like.

  Normally, the file system converts to block information that can be handled by the device driver and issues a request in block units. In the present invention, in addition to the information so far, information on the next block that is continuous as a file is also combined. Realized by issuing.

  Furthermore, this is realized by newly adding a hierarchy for managing prefetch information between the file system and the device driver. This management requires look-ahead entries and buffers, which exist independently of the cache normally managed by the file system and are shared by all paths that can reach the same disk. is there.

  Here, the prefetch entry manages block information to be prefetched and whether or not data is stored in the buffer. The buffer exists in a pair with the prefetch entry and is used to store the data of the block.

  When the request type (write or read), the target block, and the next block information are received from the file system, the prefetch information is managed in a different procedure depending on the type of request.

  The prefetch management when the request is a write is performed according to the following procedure.

  The prefetch entry is searched, and if the target block exists in the entry, the information of the entry is discarded and the prefetch information for the target block is invalidated. Furthermore, the write is executed using the operation route.

  When the request is a read, prefetch management is performed according to the following procedure.

  1. As a result of searching the prefetch entry, if the target block exists in the entry and the data is already stored in the buffer, the data in the buffer is returned.

  2. As a result of searching the prefetch entry, if the read target block exists in the entry and data is not yet stored in the buffer, the prefetch operation is being executed in the standby system. That is, after the read process being executed is completed, the data stored in the buffer is returned.

  3. As a result of searching for the prefetch entry, if the read target block does not exist in the entry, the prefetch information is not valid. That is, the read is executed using the operational route.

  4). As a result of searching the prefetch entry, if the next block to be read does not exist in the prefetch entry, it means that prefetching is not performed. That is, a prefetch entry for an external lock is created, and a read is executed using a standby path. At this time, the read data is stored in a buffer paired with the prefetched entry.

5. If data is returned from the buffer of the prefetch entry so that the above 1 and 2 correspond, the information of the prefetch entry is discarded and the entry is invalidated.
FIG. 2 shows a flow of the above procedure.

  Next, examples will be described.

  The present embodiment is realized in the connection configuration shown in FIG. The host computer 101 executes management of prefetch information in the present invention, including the operating system. The input / output unit (operation system) 102 and the input / output unit (standby system) 103 perform actual input / output in response to a read / write request to the storage device. The switch 104 and the switch 105 are necessary when the input / output unit is connected to a plurality of storage devices, and switch the connection according to the destination of the request. The storage device 106 has at least two systems of input / output, and input / output is possible in the same way from either of them.

  FIG. 3 shows details of the host computer 101 of FIG. As hardware, it has CPU110, the memory 111, and the external connection mechanism 112, and these are connected inside the host computer 101, and communication is possible between each. Further, it is connected to the input / output units 102 and 103 in FIG. On the memory 111, program information and data to be executed by the CPU 110 are held, but in the present embodiment, the following information is arranged.

  The file system area 120 is an area for the operating system to realize a function as a file system, has program information 121 to be executed by the CPU 110, and has a cache 122 and a buffer 123. The cache 122 is an area in which the file system 120 holds a value as prefetch information, and is used to hold a value that can be discarded at any time. The buffer 123 is an area for holding a value that is actually read from or written to the storage device, and is used to hold a value that cannot be discarded until reading or writing is completed.

  The prefetch function area 130 is an area for holding a function for realizing the present invention, and has program information 131 for the CPU 110 to execute the procedure of the present invention, and has a prefetch entry 132 and a buffer 133. The prefetch entry 132 holds block information and presence / absence of storage in the buffer. The buffer 133 is an area for holding a prefetched value in the storage device, and the prefetch entry 132 and the buffer 133 are managed in pairs.

  The device driver area 140 is an area for realizing an operation on the storage device via the input / output units 102 and 103, has program information 141 to be executed by the CPU 110, and holds the states of the input / output units 102 and 103. Control information 142 to be used. Further, a buffer 143 is provided as an area for holding a value to be read / written from / to the storage device.

  The operating system area 150 is an area for managing the host computer 101, has program information 151 to be executed by the CPU 110, and has control information 152 to hold the state.

  In this embodiment, the prefetch entry 132 and the buffer 133 are secured as necessary, and are released when they are no longer needed, that is, when they are invalidated. In the initial state, the prefetch entries 132 and the buffer 133 are used. None of these are present on the memory 111.

  Next, the operation of the embodiment of the best mode for carrying out the present invention will be described with reference to the drawings.

  In this embodiment, the behavior by the control of the present invention when the request is made in the order shown in FIG. 4 is shown. Further, the switching of the processing for each layer is performed as shown in FIG. 5, and a specific description will be given below.

  At time 1, a read request for the block 3 from the file system is received. At this time, since there is no prefetch entry, a read for block 3 is first issued to the operation system. Thereafter, as the prefetching process, since the next block as the file of block 3 is block 4 and there is no prefetching entry, a prefetching entry for block 4 is created and a read for block 4 is issued to the standby system.

  When the read to block 3 issued to the active system is completed, the data is returned to the file system. When the read of block 4 to the standby system executed as prefetching is completed, data is stored in the buffer corresponding to the prefetching entry, and the fact that it has been stored in the buffer is updated as information in the entry.

Only prefetch entries corresponding to block 4 exist at this time.
At time 2, a read request for the block 4 from the file system is received. At this time, since the prefetch entry for block 4 exists in the prefetch entry, priority is given to the prefetch processing to block 7 which is the next block. Since there is no prefetch entry for block 7, a prefetch entry for block 7 is created and a read for block 7 is issued to the standby system. Returning to the processing for block 4, the information of the prefetch entry is referred to, and the data in the buffer is returned to the file system since the data has already been stored in the buffer. Since the data in the buffer is returned, the prefetch entry and buffer corresponding to block 4 are discarded and invalidated.

Only prefetch entries corresponding to the block 7 exist at this time.
At time 3, a write request for the block 7 from the file system is received. At this time, since there is a prefetch entry for the block 7 in the prefetch entry, priority is given to discarding the prefetch entry. After invalidating the prefetch entry for block 7, a write for block 7 is issued to the active system. Since pre-reading is not executed at the time of writing, processing for the next block, block 8, is not performed.
There is no valid prefetch entry at this point.

  At time 4, a read request for the block 8 from the file system is received. At this time, since there is no prefetch entry, a read for block 8 is first issued to the operation system. Thereafter, as a prefetching process, a prefetching process is performed on the block 5 which is the next block. Since there is no prefetch entry, a prefetch entry for block 5 is created and a read for block 5 is issued to the standby system. When the read to the block 8 issued to the active system is completed, the data is returned to the file system. When the read to the block 5 to the standby system executed as the prefetch is completed, the data is stored in the buffer corresponding to the prefetch entry, and the fact that it has been stored in the buffer is updated as information in the entry.

  At the time when the processing is completed so far, only valid prefetch entries corresponding to the block 5 exist.

The block diagram which shows the structure of the Example of this invention. The flowchart which shows the summary of this invention. The block diagram which shows the structure of the host computer of FIG. Explanatory drawing which shows the order of a request | requirement. The flowchart which shows operation | movement of the Example of this invention.

Explanation of symbols

101 Host computer 102 Input / output unit (operational system)
103 Input / output unit (standby system)
104 switching device 105 switching device 106 storage device

Claims (1)

A computer system having a redundant connection structure, comprising a host computer, an operation system input / output unit, and a standby system input / output unit, wherein the computer has means for performing a pre-reading operation on the standby system input / output.

Images