JP6589500B2 - Information processing terminal, shared file system, shared file method, and shared file program - Google Patents

Description

  The present invention relates to an information processing terminal that can manage reading and writing of data with respect to a shared file (storage) from a client terminal side.

  In recent years, some storage servers or the like are equipped with SSDs (Solid State Drives) instead of HDDs (Hard Disc Drives) as storage devices, or both. In this case, many use high-speed and small-capacity SSDs for OS (operation system) storage, and low-speed but low-cost and large-capacity HDDs for data storage. In order to construct the fastest system possible, the fastest SSD may be used for data storage, but a large capacity SSD is expensive.

  For this reason, as a cache of a low-speed and large-capacity HDD data drive, a method using a high-speed and small-capacity SSD is used. For example, this is the optimal data placement function for storage and bcashe of Linux (registered trademark).

  However, with this method, when the distance from the cache (SSD, etc.) to the stable storage (HDD, etc.) is long, it may not operate properly as an IO (Input / Output) cache that occurs suddenly in large quantities. In addition, when there are a large number of client terminals, the storage area in the cache is quickly saturated and performance is likely to deteriorate.

  Further, when used in a system with frequent data updates, it is necessary to frequently perform cache flushing in order to maintain cache coherency between shared file systems, which causes a decrease in IO performance. As a specific example, when high-performance computing (calculation processing with a very large amount of calculation per unit time) is executed using this technique, storage that temporarily stores intermediate calculation results due to low IO performance As an area (checkpoint file), it does not show high performance.

  In order to solve this problem, a storage system that distributes processing to a plurality of client terminals has been developed. Each of the client terminals includes a cache (SDD) capable of high-speed IO.

  In such a storage system, a technique for executing IO restriction to prevent buffer data leakage during remote copy is disclosed (Patent Document 1).

  Further, in such a storage system, a technique for achieving both a thin provisioning function and high availability is disclosed (Patent Document 2).

JP 2007-18236 A JP 2009-230742 A

  In the techniques of Patent Documents 1 and 2, a pairing mechanism is provided in order to mutually use a cache between client terminals. However, a pair of client terminals is not automatically detected, and it takes time and effort for the pairing. Further, the pairing unit is a disk unit, and frequent pairing processing is necessary to realize a small-scale task.

  The present invention has been made to solve the above problems. The main object of the present invention is to provide an information processing terminal or the like that automatically performs appropriate pairing processing in a shared storage system and has high reliability at the time of cache writing.

In order to solve the above-described problem, the first feature of the present invention is to detect a terminal that can connect to a shared file system and cache data for each file unit in a cache memory area, Instruct to write to the cache memory area of the pairing device,
An information processing terminal.

A second feature of the present invention is that a plurality of the information processing terminals are provided,
A shared file system comprising: a shared file system server for sharing data files among information processing terminals.

The third feature of the present invention is to detect a terminal that can be connected to a shared file system and paired cache data for each file unit of the cache memory area,
Instructing the cache data to be written to the cache memory area of the pairing device.
It is a shared file method.

A fourth feature of the present invention is to detect a terminal that can be connected to a shared file system and paired cache data for each file unit of a cache memory area,
This is a shared file program that causes a computer to execute a function of instructing cache data to be written into a cache memory area of a terminal to be paired.

  According to the present invention, it is possible to automatically perform an appropriate pairing process in a shared storage system and provide an information processing terminal or the like having high reliability at the time of cache writing.

It is a figure which shows the structural example of the shared file system which concerns on the 1st Embodiment of this invention. It is a figure which shows the data structural example of cache statistics information. It is a figure which shows the data structural example of the cache statistics information in a cache statistics information storage part. It is a figure showing the directory structure of a cache data file. It is a figure showing the structure of the cache data file in a cache. It is a flowchart which shows operation | movement of IO request processing. It is a flowchart which shows the operation | movement of the write-in process to a shared file. It is a flowchart which shows operation | movement of the transmission process of cache statistics information. It is a flowchart which shows operation | movement of the collection process of cache statistics information. It is a flowchart which shows the operation | movement of the pairing process of the cache of a transmission side. It is a flowchart which shows the operation | movement of the pairing process of the cache by the side of reception. It is a flowchart which shows the operation | movement of the write-in process to a pairing destination. It is a flowchart which shows operation | movement of the invalidation process of cache data. It is a flowchart which shows operation | movement of the flush process of a cache. It is a flowchart which shows operation | movement of the cache pairing cancellation | release process. It is a flowchart which shows the operation | movement of the pairing cancellation | release process of the cache in the receiving side. It is a flowchart which shows the operation | movement of a file reading process. It is a flowchart which shows operation | movement of a forced cache flush process. It is a flowchart which shows the operation | movement of a file close process. It is a figure which shows the structural example of the IO terminal which concerns on the 2nd Embodiment of this invention. It is a figure which shows the internal structural example of information processing apparatus.

  Next, an embodiment of the present invention will be described with reference to the drawings. In the following description of the drawings, the same or similar parts are denoted by the same or similar reference numerals. However, the drawings schematically show the configuration of the embodiment of the present invention. Furthermore, the embodiment of the present invention described below is an example, and can be appropriately changed within a range in which the essence is the same.

<First Embodiment of the Present Invention>
Many IO processes with a high burst property do not need to share data even with a shared file system (especially intermediate files). For this reason, such a file is closed in the IO terminal as much as possible, processed, and the final result is written back to the shared file system (staging, destaging, etc.). In the first embodiment of the present invention, such IO processing is automated. This is a low-latency and high-bandwidth IO process using a large-capacity and inexpensive high-speed cache memory such as SSD, that is, the IO bandwidth to the shared file system and the IO bandwidth of the local file system Can be realized by making them as equal as possible.

(Shared file system)
A shared file system 100 according to a first embodiment of the present invention will be described with reference to FIG. As shown in FIG. 1, the shared file system 100 according to the first embodiment includes a plurality of IO terminals (information processing terminals) 1, 2, 3 and a shared file system server (hereinafter “shared FS server”). 4), and the IO terminals 1 to 3 and the shared FS server 4 are connected by a network 5 so that data communication is possible.

  The shared FS server 4 has a file system function for managing data files shared between the IO terminals 1 to 3. The shared FS server 4 includes a shared file unit 4a that is a storage device that stores a shared file.

  The IO terminals 1 to 3 have the same configuration. The IO terminal 2 includes an IO cache unit 25. The IO terminal 3 includes an IO cache unit 35. In the following, for convenience, the own terminal will be described as the IO terminal 1 and the other terminals will be described as the IO terminals 2 and 3.

(IO terminal)
As shown in FIG. 1, the OS of the IO terminals 1 to 3 includes a user area (application area), a kernel area, and a hardware area. The user area is an area that can be changed by the users of the IO terminals 1 to 3 according to their specifications. The kernel area is an OS management unit provided in common to all the IO terminals 1 to 3. Specifically, the kernel area manages resources of the IO terminals 1 to 3, and includes hardware resources and user areas in the hardware area. It manages the exchange of data such as sessions and transactions with internal software components (application programs, etc.). For example, it provides a mechanism for managing the execution state of a running program, managing hardware resources, and allowing an application program to use hardware functions.

  An application unit 11, an input / output library unit 12, and an IO cache unit 15 exist in the user area. The kernel area includes an application virtual file system unit (hereinafter referred to as “AP virtual FS unit”) 13, a user file system control unit (hereinafter referred to as “user FS control unit”) 14, and an IO virtual file system unit (hereinafter referred to as “AP virtual FS unit”). , “IO virtual FS unit”) 16, local file system unit (hereinafter referred to as “local FS unit”) 161, and shared file system client control unit (hereinafter referred to as “shared FS client control unit”) 162. Exists. The SDD 17 and the network connection unit 18 exist in the hardware area.

  First, each part of the user area will be described.

  The IO cache unit 15 is a mechanism for executing IO control on the cache (SSD 17) and the shared FS server 4, and creates a cache data file for IO processing. The IO cache unit 15 includes an IO cache control unit 151 and a data flow rate control unit 152.

  The IO cache control unit 151 includes a pairing unit 154. The pairing unit 154 selects a pairing destination IO terminal for providing a copy of the cache data file, and performs pairing processing, pairing cancellation processing, and the like with the selected IO terminal.

  The IO cache control unit 151 can access the network connection unit 18 and performs data communication with the IO cache units 25 and 35 included in the other IO terminals 2 and 3 via the network 5.

  The IO cache control unit 151 is connected to the data flow rate control unit 152.

  The data flow rate control unit 152 makes an IO request to the shared FS client control unit 162 via the IO virtual FS unit 16.

  The data flow rate control unit 152 periodically notifies the other IO terminals 2 and 3 of the cache statistics information of itself (IO terminal 1) via the network connection unit 18. The notified cache statistical information is stored in the cache statistical information storage unit 153.

  In addition, the data flow rate control unit 152 collects cache statistical information from the other IO terminals 2 and 3 via the network connection unit 18, and stores the collected cache statistical information in the cache statistical information storage unit 153.

  As shown in FIG. 2, the cache statistical information issued by the own IO terminal 1 and the other IO terminals 2 and 3 includes “IP (internet protocol) address”, “cache free capacity”, “shared FS data flow rate”, “update”. "Time" is included as a data item. The “IP address” is an IP address that can identify the IO terminals 1 to 3. “Cache free capacity” indicates the free capacity of the SSD 17 of the IO terminals 1 to 3 having the IP address. “Shared FS data flow rate” is the number of I / O accesses (IOPS) that can be processed by the shared FS server 4 per second, and the data amount per second (MByte / s) and measured values. The “update time” is a time stamp of the time when the data update process is performed in the shared FS server 4.

  As shown in FIG. 3, the cache statistical information storage unit 153 stores cache statistical information of the other IO terminals 2 and 3 together with cache statistical information of the own IO terminal 1. The IO terminals 1 to 3 are determined using “IP address” as an identifier.

  The application unit 11 stores an application program executed by the user. The application program is appropriately calculated and executed by a CPU (Central Processing Unit) (not shown). Details will be described later.

  In many cases, the application program includes an IO process (file open process and close process) to the cache data file (SDD 17) and the shared file unit 4a.

  The input / output library unit 12 is a data file that stores functions necessary for executing the IO processing included in the application program.

  Next, each part of the kernel area will be described.

  The IO virtual FS unit 16 is a virtual file system for executing IO processing. The IO virtual FS unit 16 is connected to the local FS unit 161 and the shared FS client control unit 162, and uses these transparently.

  The SSD 17 is directly connected to the local FS unit 161. This enables high-speed and low-latency storage access.

  The shared FS client control unit 162 performs data communication with the shared FS server 4 on the network 5 via the network connection unit 18.

  The local FS unit 161 constructs a data file directory inside the cache (SSD 17) and holds the directory structure. An example of the directory structure is shown in FIG.

  In the directory structure shown in FIG. 4, the root directory is a root, and there are a directory a, a directory b,..., A directory N, and N subdirectories. Each subdirectory includes a cache data file. The directory N for certain data is determined as one of 1 to N using a hash function f (x) (x = local FS file handle value) in order to avoid access concentration.

  The data file structure inside the cache (SSD 17) is shown in FIG. The data file structure includes “pair identifier”, “master backup flag”, “block number”, “cache data”, “update flag”, and “update time” as data items.

  The “pair identifier” is an identifier (pair identifier) that can identify a pairing destination, for example, an IP (Internet Protocol) address. The “master / backup flag” is a flag indicating whether data associated with the pair identifier is master (held by the own IO terminal) data or backup (held as a pairing destination) data. “Block number” is a block number in which cache data is stored. “Cache data” is the data body cached in the SSD 17.

  The “update flag” is a flag indicating a data state. The update flags include “MODIFIED (update state: data is updated latest)”, “SHARED (shared: synchronized with the data of the shared file part 4a in the shared FS server 17)”, “INVALID (invalid ) ”And“ NO FILE (no data exists) ”. “Update time” is a time stamp of the updated time.

  The shared FS client control unit 162 is a mechanism for controlling the right to access the file system of the shared FS server 4.

  The AP virtual FS unit 13 is a virtual file system used when an application is used.

  The user FS control unit 14 is a mechanism (framework) for mounting a file system as a program that operates in the user space, and specifically, FUSE (Filesystem in USErspace) or the like.

  Next, each part of the hardware area will be described.

  The SSD 17 is used as a cache memory on the IO terminals 1 to 3 side, which is a semiconductor memory capable of high-speed IO.

  The network connection unit 18 is an interface for communication-connecting the own IO terminal 1 to the other IO terminals 2 and 3 and the shared FS server 4 via the network 5.

(Operation of shared file system)
Next, operations of the IO terminals 1 to 3 in the shared file system 100 will be described. The main operations of the IO terminals 1 to 3 are: (1) IO request processing, (2) write processing to a shared file, (3) transmission and collection processing of cache statistical information, and (4) cache pairing processing. , (5) Pairing destination write processing, (6) Cache data invalidation processing, (7) Cache flush processing, (8) Cache pairing release processing, (9) File read processing, (10 ) There is a file close process. Hereinafter, these processes will be described in detail with reference to the drawings.

(1. IO request processing)
The IO request processing will be described with reference to the flowchart of FIG.

  First, in step A1, when an application program included in the application unit 11 is executed by a CPU (not shown), an IO request process included in the description of the program is read. As a result, the CPU performs an open process of a data file that requires execution of the program. The CPU refers to the open function in the input / output library unit 12 and calls the open system call.

  In step A2, the AP virtual FS unit 13 selects a file system using the path name of the data file to be opened. At this time, when an access path for the cache file system is designated, the IO cache unit 15 mounted in the user space is called through the user FS control unit 14.

  In step A3, the IO cache control unit 151 of the IO cache unit 15 creates a cache data file, and causes the IO virtual FS unit 16 to select the local FS unit 161 at that time.

  In step A <b> 4, the IO virtual FS unit 16 creates a cache data file for storing cache data in the SSD 17. At this time, the IO virtual FS unit 16 returns a file handle identifier for uniquely identifying the created file to the AP virtual FS unit 13 via the IO cache unit 15.

  This completes the IO request processing.

(2. Write processing to shared file)
The writing process to the shared file will be described with reference to the flowchart of FIG.

  In step B1, according to the description of the application program stored in the application unit 11, the CPU refers to the write function in the input / output library unit 12 and calls the write system call.

  In step B2, the AP virtual FS unit 13 selects a file system based on the file handle identifier acquired by the above open process.

  In step B3, the AP virtual FS unit 13 determines that the IO cache unit 15 installed in the user area via the user FS control unit 14 when the file handle identifier indicates a file handle for the cache file system. Call.

  In step B <b> 3, the pairing unit 154 included in the IO cache control unit 151 of the IO cache unit 15 checks whether there are other IO terminals 2 and 3 already paired in the own IO terminal 1. If there are other IO terminals 2 and 3 that have already been paired, the process proceeds to step B5. If there is no other IO terminal 2 or 3 that has already been paired, the process proceeds to step B4.

  In step B <b> 4, the pairing unit 154 performs pairing processing between the own IO terminal 1 and the other IO terminals 2 and 3. Details of the pairing process will be described later.

  In step B5, the IO virtual FS unit 16 requests the IO cache unit 15 to write the cache data file in order to write the cache data file to the local SSD 17. The IO cache control unit 151 of the IO cache unit 15 writes data to the SSD 17 via the IO virtual FS unit 16 and the local FS unit 161. The data format to be written is executed in the format of the data file structure shown in FIG. If the write is an update to the existing cache data in the SSD 17, the block update flag of the cache data file (see FIG. 5) is set to “MODIFIED (update state: data is updated to the latest)”. Change to

  In step B6, simultaneously with the writing process of the cache data file, the IO cache unit 15 refers to the pair identifier of the cache data file and refers to the other IO terminals 2 and 3 (pairing destination) corresponding to the pair identifier. However, the same cache data is transmitted, and the writing is requested to the paired destination SDD 17 as well. Details of the writing process at the pairing destination will be described later.

  In step B7, the IO cache control unit 151 of the IO cache unit 15 updates the cache statistical information (see FIG. 2) of the own IO terminal 1.

  In step B8, the IO cache control unit 151 confirms the completion of the pairing destination write process (synchronization of write). Completion of the write process is confirmed by receiving a write process completion notification from the pairing destination. In the case of a program with many update processes, the program may be determined after a predetermined time.

  Thus, the writing process to the shared file is finished.

(3. Sending and collecting cache statistics information)
Cache statistics information transmission processing will be described with reference to the flowchart of FIG.

  In step C1, the IO cache control unit 151 of the IO cache unit 15 creates cache statistical information (see FIG. 2) of the own IO terminal 1, and broadcasts it to the network 5 via the network connection unit 18.

  In step C2, the IO cache control unit 151 then waits for a predetermined time and confirms the elapse of the predetermined time (step C3). Then, the cache statistics information of the own IO terminal 1 is created again, via the network connection unit 18 And broadcast transmission to the network 5.

  Thus, the cache statistical information transmission process is completed.

  Next, cache statistics information collection processing will be described with reference to the flowchart of FIG.

  In step C4, the IO cache unit 15 of the own IO terminal 1 is in a standby state until it receives the cache statistical information from the other IO terminals 2 and 3. When the cache statistical information is received, the process proceeds to Step C5. If the cache statistical information is not received, the standby state in step C4 is continued.

  In step C5, the IO cache unit 15 acquires cache statistical information from the other IO terminals 2 and 3.

  In step C6, the IO cache unit 15 updates the cache statistical information table (see FIG. 3) in the cache statistical information storage unit 153 with the acquired cache statistical information. The cache statistical information in the cache statistical information storage unit 153 is used in the pairing destination determination process.

  Thus, the cache statistical information collection process is terminated.

(4. Cache pairing process)
The cache pairing process on the transmission side (the own IO terminal 1 in the following description) will be described with reference to the flowchart of FIG.

  In Step D1, the IO cache unit 15 of the own IO terminal 1 that is the transmission side of the pairing process searches the cache statistical information in the cache statistical information storage unit 153 and selects a pairing destination. Specifically, the pairing unit 154 of the IO cache control unit 151 requests the data flow rate control unit 152 to select an optimal pairing destination. Upon receiving the request, the data flow rate control unit 152 refers to the cache statistical information (FIG. 3) of the other IO terminals 2 and 3 stored in the cache statistical information storage unit 153, and has a large cache free capacity and a shared file. An IO terminal with a small system data flow rate value (here, IO terminal 2) is selected.

  In step D <b> 2, the pairing unit 154 transmits a pairing processing request to the selected IO terminal 2. In step D3, the pairing unit 154 that has received the result of the pairing processing request from the selected IO terminal 2 determines whether the result is permission for pairing acceptance or disapproval for acceptance (step D4). If the result is permission to accept pairing, the process proceeds to step D5. If the result is that pairing acceptance is not permitted, the process returns to step D1, and another pairing candidate IO terminal is reselected.

  In step D5, the pairing unit 154 temporarily stores the identifier (IP address) of the IO terminal 2 that is the pairing destination, assuming that pairing has been established. You may store as a pair identifier of the cache data file shown in FIG. After pairing is established, there is no need to examine the data transmission source when transmitting and receiving cache data.

  This completes the cache pairing process on the transmission side.

  Next, cache pairing processing on the receiving side (in the following description, the other IO terminal 2) will be described with reference to the flowchart of FIG.

  In step D6, the IO terminal 2 on the receiving side of the pairing process determines whether or not a pairing request has been received in the standby state for receiving a pairing request. When the pairing request is received, the process proceeds to Step D7. If no pairing request is received, the standby state is continued in step D6.

  In step D7, the pairing unit 154 of the IO cache control unit 151 refers to the cache statistical information storage unit 153, and based on the cache statistical information of itself (that is, the IO terminal 2), other IO terminals that are paired Investigate whether or not there is sufficient cache capacity.

  In step D8, based on the result of the investigation, the pairing unit 154 determines whether or not pairing acceptance can be permitted. If the acceptance of pairing is permitted, the process proceeds to step D10. If the pairing acceptance is not permitted, the process proceeds to Step D9.

  In step D9, the pairing unit 154 transmits a pairing acceptance non-permission notification to the IO terminal 1 that is the pairing request source.

  In step D10, the pairing unit 154 transmits a pairing acceptance permission notification to the IO terminal 1 that is the pairing request source.

  In step D11, the pairing unit 154 opens the pairing cache file. This open process is executed from the cache data file creation process in step A3 in the IO request process (see FIG. 6). Further, the pairing unit 154 temporarily stores the IP address of the IO terminal 1 that is the pairing source and the time (reference time) when the cache data file was opened. You may store as a pair identifier of the cache data file shown in FIG.

  This completes the cache pairing process on the receiving side.

(5. Write processing to pairing destination)
Next, the writing process to the pairing destination will be described with reference to the flowchart of FIG.

  In step D21, the IO terminal 2 that is the pairing destination of the IO terminal 1 takes a standby state for cache data. When the cache data is received, the process proceeds to Step D22. While the cache data is not received, the standby state of D21 is continued.

  In step D22, the IO cache control unit 151 determines whether or not the transmission source of the cache data is the IO terminal 1 that has permitted pairing. Specifically, the IO cache control unit 151 compares the transmission source IP address with the IP address temporarily stored in the pairing unit 154, and determines whether or not they match. If it is determined that the transmission source is an IO terminal that permits pairing, the process proceeds to step D24. If it is determined that the transmission source is not an IO terminal that permits pairing, the process proceeds to D23.

  In step D23, the IO cache control unit 151 transmits a write disable notification to the cache data transmission source via the network connection unit 18.

  In step D24, the IO cache control unit 151 uses the pair identifier (see FIG. 5) of the file that has been opened for the IO terminal 1 that has permitted pairing in the receiving-side pairing process (see step D11 in FIG. 11). The cache data file for pairing is searched for, and the cache data is written into the searched cache data file (SSD 17).

  In step D <b> 25, when the writing of the cache data is completed, the IO cache control unit 151 transmits a write completion notification to the cache data transmission source IO terminal 1 via the network connection unit 18.

  In step D <b> 26, the IO cache control unit 151 broadcasts the updated cache block metadata to the network 5 via the network connection unit 18.

  Thus, the writing process to the pairing destination is completed.

  Note that the metadata broadcast transmission is performed in order to inform all the IO terminals on the network 5 that the cache data of the pairing destination (IO terminal 1) is held by itself (IO terminal 2). The metadata includes an identifier of the own IO terminal 2, an identifier of the pairing destination IO terminal 1, and a range (block number, number of blocks, etc.) holding cache data.

  Also, the other IO terminal (for example, IO terminal 3) that has received the metadata checks whether it holds the same cache data file in its SSD 17, and if so, the corresponding block in the cache data file Change the update flag to “INVALID”. This is to prevent the other IO terminal 3 from holding an unnecessary cache data file and preventing the other IO terminal 3 from erroneously transmitting this data to an upper application or the like.

(6. Cache data invalidation processing)
In order to maintain consistency, the cache data needs to be held only by the master IO terminal 1 and the backup IO terminal 2 after pairing. When the cache data is held by an IO terminal other than the master side and the backup side (referred to as IO terminal 3), it is necessary to invalidate the data so that it is not returned to an upper application on the network 5. There is.

  The cache data invalidation process will be described with reference to the flowchart of FIG.

  In step E1, the IO cache unit 15 of the IO terminal (referred to as the IO terminal 2) determines whether or not the updated cache block metadata has been received. If metadata is received, the process proceeds to step E2. While the metadata is not received, the process returns to step E1 and the standby state is continued.

  In step E2, based on the received metadata, the cache statistics information storage unit 153 is examined to determine whether the range of the cache data described in the metadata is held in the own IO terminal 2. As a result of the investigation, when the range of the cache data described in the metadata is held in the own IO terminal 2, the process proceeds to step E3. If the own IO terminal 2 does not hold it, this process is terminated once and the process returns to the standby state of step E1 again.

  In step E3, it is determined whether or not the transmission source (IP address) of the cache data is a master IO terminal (referred to as IO terminal 1) that is a pairing destination. This is because if the transmission source of the cache data is the pairing destination, it is the cache data being written, and it can be determined that no problem occurs. In this case, the process is once ended and the process returns to the standby state of Step E1 again. If the transmission source of the cache data is not the pairing destination, the process proceeds to Step E4.

  In step E4, the IO cache unit 15 invalidates the cache data of the unpaired IO terminal (for example, the IO terminal 3) stored in the cache statistics information storage unit 153. Specifically, the IO cache unit 15 changes the update flag of the corresponding cache data in the cache data file (see FIG. 5) to “INVALID (invalid)”. This process is performed in all corresponding blocks in the cache data file.

  Thus, the cache data invalidation process is completed.

(7. Cache flush processing)
Next, cache flush processing will be described with reference to the flowchart of FIG. Note that the cache flush process operates as appropriate (for example, at predetermined intervals) as an independent process on the IO terminal 1.

  In step F1, the IO cache unit 15 searches for an updated block. Specifically, the IO cache unit 15 searches for a block in which the update flag of the cache file data (see FIG. 5) is “MODIFIED (update state: data is updated to the latest)”, and cache flush target Select as a block.

In step F2, the data flow rate control unit 152 of the IO cache unit 15 uses the MRU (Maximum Receive unit) algorithm to generate cache statistical information having the same IP address as the pair identifier (see FIG. 5) of the cache flush target block. The shared FS data flow rate is confirmed from the record of FIG. 3), and it is confirmed whether an IO request exceeding a predetermined threshold value is issued (step F3). If IO requests exceeds a predetermined threshold value is out, the process proceeds to step F 4. If IO requests exceeds a predetermined threshold value is not out, the process proceeds to step F 6.

  In step F4, the data flow rate control unit 152 determines whether or not the cache free capacity is equal to or less than a predetermined threshold value. If the cache free space is equal to or smaller than the predetermined threshold, the process proceeds to Step F6. If the cache free capacity is not less than the predetermined threshold value, the process proceeds to F5.

  In step F <b> 6, the IO cache control unit 151 of the IO cache unit 15 passes the cache data to the shared FS server 4 via the network connection unit 18. The shared FS server 4 writes the cache data in the shared file unit 4a. At this time, the IO cache control unit 151 sets the write time to the “update time” of the cache data file (see FIG. 5) in the SSD 17 via the IO virtual FS unit 16, and sets the “update flag”. Set to “SHARED (shared: synchronized with data of shared file unit 4a in shared FS server 17)”. This means that the cache (SSD 17) and the shared file server 4 hold the same data.

  In step F5, if there is no IO request exceeding the threshold value and the cache free capacity is still sufficient, or if the flush process to the shared FS server is completed, after waiting for a predetermined time, the process proceeds to the remaining data amount in step F2. Returned to the confirmation process.

  Thus, the cache flush process is completed.

(8. Cache pairing release processing)
Next, processing in the transmitting-side IO terminal that requests cache pairing cancellation (which will be described as the IO terminal 2 in the following description) will be described with reference to the flowchart of FIG.

  First, in step G1, the IO cache unit 15 of the own IO terminal 2 takes a standby state for a predetermined time. This standby state is a time during which it is possible to determine that the IO processing congestion that has become a bottleneck during the operation of a certain application program has been alleviated by the pairing process, or that the application program has ended. ~ Several hours etc. When the standby state for a certain period of time ends, the process proceeds to Step G2.

  In step G2, the IO cache unit 15 acquires the reference time of the currently opened cache data file from the pairing unit 154 (see step D11 in FIG. 11).

  In step G3, the IO cache control unit 151 of the IO cache unit 15 compares the acquired reference time with the current time acquired from the system, and determines whether a preset specified time has elapsed. If the specified time has elapsed, the process proceeds to step G4. If the specified time has not elapsed, the process returns to G1.

  In step G4, the pairing unit 154 of the IO cache control unit 151 transmits a pairing release request to the IO terminal 1 that is the pairing destination.

  In step G5, the pairing unit 154 waits for a pairing release completion notification from the paired IO terminal 1.

  In step G6, the pairing unit 154 performs a close process for the cache data file after receiving the notification of the completion of the pairing release.

  This completes the processing at the transmitting IO terminal 2 of the cache pairing release request.

  Next, the processing in the receiving IO terminal (hereinafter described as the IO terminal 1) for releasing the cache pairing will be described with reference to the flowchart of FIG.

  In step G7, the receiving-side IO terminal 1 of the pairing cancellation process is in a waiting state for receiving a pairing cancellation request. In this state, when a pairing release request is received, the process proceeds to step G8. The reception waiting state in step G7 is continued until a pairing release request is received.

  In step G8, when a pairing release request is received, the receiving side IO terminal 1 of the pairing release process performs a process of closing the cache data file that is currently open.

  In step G9, when the closing process is completed, the pairing unit 154 returns a pairing cancellation completion notification to the transmission-side IO terminal 2 of the cancellation request.

  Thus, the processing in the receiving IO terminal 1 for canceling the cache pairing is completed.

(9. File reading process)
Next, the file reading process will be described with reference to the flowchart of FIG.

  In step H1, the CPU calls a read system call in the read function provided by the input / output library unit 12 based on the description of the application program in the application unit 11 (see FIG. 1).

  In step H2, the AP virtual FS unit 13 selects a file system based on the file handle identifier (see step A4 in FIG. 6) acquired by the open process. At this time, when the file handle identifier indicates the file handle for the cache FS, the AP virtual FS unit 13 passes through the user FS control unit 14 and the IO cache unit 15 mounted on the user area. Call.

  In step H3, the IO cache control unit 151 of the IO cache unit 15 searches the cache file on the local FS unit 161 and checks the update flag of each block in the cache file. At this time, the update flag becomes “MODIFIED (update state: data is updated latest)” or “SHARED (shared: synchronized with the data of the shared file unit 4a in the shared FS server 17)”. If so, the cache data is returned to the upper application as it is (step H7). If the update flag is “NO FILE (no cache file exists)” or “INVALID (invalid)”, the process proceeds to step H4.

  In step H4, the IO cache unit 15 issues a forced flush request to all other IO terminals 2 and 3 in order to obtain valid data. First, the IO cache unit 15 broadcasts the metadata of the block whose update flag is “NO FILE” or “INVALID (invalid)” via the network connection unit 18. This metadata includes a cache file name, a block number, the number of blocks, and the like. When all the other IO terminals 2 and 3 receive this flush request, the local IO terminals 2 and 3 perform forced flush processing. Details of this processing will be described later.

  In step H5, when the forced flush process ends and the desired data is updated to the shared FS server 4, the IO cache unit 15 sends the forced flushed desired data to the shared FS via the network connection unit 18. Read from server 4 and get.

  In step H6, the IO virtual FS unit 16 writes the acquired data to the SSD 17 as cache data.

  Next, the forced flush process in step H4 will be described with reference to the flowchart of FIG.

  In step H8, the IO cache unit 15 of the IO terminals 2 and 3 receives the flush request. This flush request is metadata or data including metadata.

  In step H9, the IO cache control unit 151 of the IO cache unit 15 checks whether the cache file name, block, and data corresponding to the range included in the metadata exist in its own cache (SSD 17). If it exists, the process proceeds to Step H10. If not, the process returns to step H8.

  In step H <b> 10, the IO cache control unit 151 transmits the cache data requested to be flushed to the shared FS server 4 via the network connection unit 18. The shared FS server 4 writes the cache data requested for flushing to the shared file unit 4a. At this time, the update time in the cache data file is set.

  In step H11, the IO cache control unit 151 sets “SHARED” to the update flag of the corresponding block in the cache file data (FIG. 5).

  Thus, the forced flush process is completed.

(10. File closing process)
Next, file closing processing for ending an IO request by the application unit 11 (see FIG. 1) will be described with reference to the flowchart of FIG.

  In step I1, according to an application program stored in the application unit 11, a CPU (not shown) calls a close system call in a close function provided by the input / output library unit 12.

  In step I2, the AP virtual FS unit 13 receives a file handle value of a file to be closed and selects a file system.

  When the file handle for the cache FS is received in step I3, the AP virtual FS unit 13 calls the IO cache unit 15 mounted on the user area via the user FS control unit 14. The IO cache unit 15 closes the cache data file.

  In step I4, the IO virtual FS unit 16 selects the local FS unit 161, and the local FS unit 161 deletes the cache data file created on the SSD 17.

  As described above, the file closing process for ending the IO request is ended.

(effect)
According to the first embodiment of the present invention, in the shared storage system, appropriate pairing processing is automatically performed, and the IO terminals 1 to 3 for using the shared file system with high reliability at the time of cache writing are used. Can be provided.

  This is because the IO terminal as a pairing destination for temporarily storing a copy of its own cache data is considered for each file unit block in consideration of the cache usage amount of the IO terminal and the access amount to the shared FS server 4. It is because it chooses. Thereby, the reliability at the time of writing can be made high. Furthermore, the original IO request of the shared file system can be prevented from being hindered.

<Second Embodiment>
FIG. 20 shows an IO terminal (information processing terminal) 10 for using the shared file system according to the second embodiment of the present invention. The IO terminal 10 detects a terminal for which cache data is paired for each file unit in the cache memory area, and instructs to write the cache data in the cache memory area of the paired terminal.

  According to the second embodiment of the present invention, in the shared storage system, it is possible to automatically perform an appropriate pairing process and provide the IO terminal 10 with high reliability at the time of cache writing.

  This is because the IO terminal as a pairing destination for temporarily storing a copy of its own cache data is considered for each file unit block in consideration of the cache usage amount of the IO terminal and the access amount to the shared file system server. It is because it chooses.

<Example of change>
In the above embodiment, FUSE is used as the user file system control unit 14 and the file system is configured by implementation in the user area. However, the same function can be achieved by implementation using a kernel module or implementation in the kernel. It is feasible.

  In order to maintain cache consistency, it is desirable for cache control to use the MSI (Modified / Shared / Invalid) protocol for simplicity, but other cache control algorithms may be applied.

  In the above embodiment, only one IO terminal that has a cache copy is selected, but two or more IO terminals may be selected to increase redundancy.

(Hardware configuration)
In each embodiment of the present invention, each component in the IO terminals 1 to 3 represents a functional unit block. Some or all of the components in the IO terminals 1 to 3 are realized by any combination of an information processing apparatus and a program as shown in FIGS. An example of the information processing apparatus 500 that realizes this is shown in FIG. The information processing apparatus 500 includes the following configuration as an example.

CPU 501
ROM 502
RAM 503
A program 504 loaded into the RAM 503
A storage device 505 for storing the program 504
A drive device 507 for reading / writing the recording medium 506
Communication interface 508 connected to the communication network 509
An input / output interface 510 for inputting / outputting data
-Bus 511 connecting each component
Each component of the IO terminals 1 to 3 in each embodiment is realized by the CPU 501 acquiring and executing a program 504 that realizes these functions. The program 504 that realizes the function of each component of each device is stored in advance in the storage device 505 or the RAM 503, for example, and is read by the CPU 501 as necessary. The program 504 may be supplied to the CPU 501 via the communication network 509 or may be stored in the recording medium 506 in advance, and the drive device 507 may read the program and supply it to the CPU 501.

  There are various modifications to the IO terminal implementation method. For example, the IO terminal may be realized by an arbitrary combination of the information processing apparatus 500 and a program that are separately provided for each component. In addition, a plurality of components included in the IO terminal may be realized by any combination of one information processing apparatus 500 and a program.

  Also, some or all of the components of the IO terminal are realized by other general-purpose or dedicated circuits, processors, etc., or combinations thereof. These may be configured by a single chip or may be configured by a plurality of chips connected via a bus.

  Some or all of the components of the IO terminal may be realized by a combination of the above-described circuit and the like and a program.

  When some or all of the components of the IO terminal are realized by a plurality of information processing devices and circuits, the plurality of information processing devices and circuits may be centrally arranged or distributedly arranged. Also good. For example, the information processing apparatus, the circuit, and the like may be realized as a form in which each is connected via a communication network, such as a client and server system and a cloud computing system.

  Each embodiment mentioned above is only an example about application of the technical idea obtained by this inventor. That is, the technical idea is not limited to the above-described embodiment, and various changes can be made without departing from the spirit of the invention.

1: IO terminal 2: IO terminal 3: IO terminal 10: IO terminal 4: Shared FS server 4a: Shared file unit 5: Network 11: Application unit 12: Input / output library unit 13: AP virtual FS unit 14: User FS control Unit 15: IO cache unit 16: IO virtual FS unit 17: shared FS server 18: network connection unit 25: IO cache unit 35: IO cache unit 100: shared file system 151: IO cache control unit 152: data flow rate control unit 153 : Cache statistical information storage unit 154: Pairing unit 161: Local FS unit 162: Shared FS client control unit 500: Information processing device 501: CPU
502: ROM
503: RAM
504: Program 505: Storage device 506: Recording medium 507: Drive device 508: Communication interface 509: Communication network 510: Input / output interface 511: Bus

Claims (7)

Connectable to a shared file system, detects the terminal to pair cache data for each file unit of the cache memory area, the cache data, and it instructs to write the cache memory area of the terminal of the pairing,
Broadcast information to the external network indicating the free capacity of the cache memory area of its own terminal,
Information processing terminal. When the cache data is received for each file unit from the pairing terminal, the cache data is written in the cache memory area of the own terminal.
The information processing terminal according to claim 1. The information processing terminal according to claim 1 or 2 , wherein information indicating a free capacity of a cache memory area of another terminal is received from the external network, and the terminal to be paired is determined based on the information. Detecting the data output amount from the information processing terminal to the shared file system server, if the data input amount exceeds a predetermined threshold, adjusting the input and output of data to the shared file system server from the information processing terminal,
The information processing terminal according to any one of claims 1 to 3. A plurality of information processing terminals according to any one of claims 1 to 4,
A shared file system server for each of the information processing terminals to share a data file;
A shared file system comprising: Connectable to a shared file system, detects the terminal to pair cache data for each file unit of the cache memory area,
The cache data, and instructs to write the cache memory area of the terminal of the pairing,
Broadcast information to the external network indicating the free capacity of the cache memory area of its own terminal,
Shared file method. Connectable to a shared file system, detects the terminal to pair cache data for each file unit of the cache memory area,
The cache data, and instructs to write the cache memory area of the terminal of the pairing,
Broadcast information to the external network indicating the free capacity of the cache memory area of its own terminal,
A shared file program that causes a computer to execute a function.

Images