JP6214089B2 - Computer system and program for determining data replication destination - Google Patents

Description

The present invention relates to a computer system and a program for determining a storage as a data replication destination.

  As a method of data protection, Non-Patent Document 1 is based on a storage system installed at geographically dispersed sites, and quantitatively evaluates the risk of data loss against disasters such as tsunamis, and finds low-risk sites. A method of selectively replicating data to a secure site is proposed.

“Proposal of a risk-aware data replication method for wide-area disasters” IEEJ 55th Information Systems Research Materials, pages 39-43, 2013.6.21

  The technology as shown in Non-Patent Document 1 explosively increases the time required to determine the replication destination because the number of node combinations increases when the number of storage systems and the number of replications at each site constituting the storage system increases. There was a problem to do.

  Therefore, a computer system that can determine the data replication destination in a short time is required.

  The disclosed computer system has a plurality of storages and a management server connected to the plurality of storages via a network. The management server uses one of the plurality of storages that stores the data replication data by using the data loss weight of the data corresponding to the replication source storage of each of the plurality of storages storing the data. The data is determined as the first replication destination storage, the data loss weight is updated using the given disaster risk, and the replication data is stored in correspondence with the replication source storage storing the data using the updated data loss weight. Another one of the plurality of storages to be stored is determined as the second replication destination storage.

  According to the present invention, a computer system capable of determining a replication destination in a short time is realized.

It is a block diagram of a computer system. It is a block diagram of a storage. It is a structural example of a storage configuration management table. It is a structural example of a disaster risk management table. It is an example of a structure of a data loss weight management table. It is an example of a structure of a replication structure management table. It is a flowchart of the replication destination determination process which determines a 1st replication destination. It is a structural example of a combination work table. It is a flowchart of the combination process which determines the combination of a replication source and a replication destination. It is an example of composition of an exchange work table. It is a flowchart of the combination replacement | exchange process which shows the procedure switched to a better combination. It is a flowchart of the replication destination determination process which determines a 2nd replication destination. It is an example of a structure of the data loss weight management table which updated the data loss weight. It is a flowchart of the data loss weight update process which updates a data loss weight. It is a figure which shows the effect by Example 1. FIG. It is a structural example of a data redundancy management screen. It is an example of a structure of the data loss weight management table which updated the data loss weight in Example 2. FIG. 10 is a flowchart illustrating a procedure for updating data loss weights in the second embodiment.

  Hereinafter, embodiments will be described with reference to the drawings. In the following description, various types of information may be described using an expression such as “management table”, but the various types of information may be expressed using a data structure other than a table. Further, the “management table” can be referred to as “management information” to indicate that it does not depend on the data structure.

  Further, the process may be described with “program” as the subject. The program is executed by a processor, for example, an MP (Micro Processor) or a CPU (Central Processing Unit), and performs a predetermined process. Note that the subject of processing may be a processor because the storage resource (for example, a memory) and a communication interface device (for example, a communication port) are used as appropriate. The processor may have dedicated hardware in addition to the CPU. The computer program may be installed on each computer from a program source. The program source may be provided by, for example, a program distribution server or a storage medium.

  Each element such as a table can be identified by a number or the like, but other types of identification information such as a name may be used as long as the information can be identified. In the drawings and description of the present invention, the same reference numerals are given to the same parts, but the present invention is not limited to the present embodiment, and any application examples that meet the idea of the present invention are technical. Included in the range. Further, unless specifically limited, each component may be plural or singular.

  In the present embodiment, an iterative determination of a data replication destination accompanied by an update of data loss weight will be described.

  FIG. 1 is a configuration diagram of a computer system 100 according to this embodiment. The computer system 100 is configured by connecting a plurality of storages 101 (storages 101 a to 101 n) and a management server 102 via an operation network 105. The management server 102 includes a CPU 106, a memory 107, an interface 108 (hereinafter referred to as an operation I / F 108) for connecting to the operation network 105, and an interface 109 (hereinafter referred to as a console I / F 109) for connecting to the management console 103. Have.

  The memory 107 includes a storage configuration management table 110, a disaster risk management table 111, a data loss weight management table 112, and a replication configuration management table 113, a data loss weight calculation program 114, a replication destination determination program 115, and Each program (processing unit) of the replication control program 116 is stored, and the CPU 106 executes these programs.

  The operation network 105 is a network for the management server 102 to operate the storage 101, and Ethernet (registered trademark) or the like is preferable. The data network 104 is a network for transferring data between the plurality of storages 101, and Ethernet, fiber channel, the Internet, etc. are preferable. Note that the data network 104 may be the same network as the operation network 105.

  FIG. 2 is a configuration diagram of the storage 101 of this embodiment. The storage 101 includes a CPU 200, a memory 201, a volume 202 for storing data, an interface 203 (operation I / F 203) for connecting to the operation network 105, and an interface 204 (data I / F 204) for connecting to the data network 104. ).

  The memory 201 stores a replication program 205, and the CPU 200 executes the replication program 205. The replication program 205 replicates the data stored in the volume 202 to the other storage 101 via the replication program 205 of the other storage 101 and the data network 104. Data duplication units are block units, file units, object units, and the like.

  The table and each program stored in the memory 107 may be stored in the memory 201, and the CPU 200 may execute each program.

  Hereinafter, a case where the computer system 100 is connected to four storages 101 and the data redundancy of each storage 101 is set to 3 (that is, when two copies of data stored in each storage are created) is taken as an example explain.

  FIG. 3 is a diagram illustrating a configuration example of the storage configuration management table 110. The storage configuration management table 110 includes an ID 300 that identifies the storage 101, an amount of data stored in each storage 101 (amount of original data to be replicated) 301, a free capacity 302 in which each storage 101 stores replicated data, It has an IP address 303 necessary for accessing the storage 101. Each of the data amount 301 and the free capacity 302 may indicate an absolute amount in units of MB or the like, but here, a value standardized with a predetermined data amount is used for simplification.

  The storage configuration management table 110 is created by an administrator using the management console 103 prior to data replication.

  The IP address 303 is shown as an example of the IP address of the operation I / F 203 of each storage 101, but may be, for example, the data I / F 204 or both depending on the network configuration. If the network is configured with a fiber channel, it may include a WWN.

  FIG. 4 is a diagram illustrating a configuration example of the disaster risk management table 111. The disaster risk management table 111 stores the values of the disaster risk 402 for each combination of the replication source (replication source storage) 400 and the replication destination (replication destination storage) 401. The disaster risk 402 can be regarded as a disaster risk corresponding to the data when the data is copied from the copy source (copy source storage) 400 to the copy destination (copy destination storage) 401.

  The disaster risk is, for example, the probability that the storage 101 of the copy source 400 and the copy destination 401 is damaged at the same time due to a disaster such as an earthquake or a tsunami, and the stored data of the storage 101 is lost.

  In the example shown in FIG. 4, when data is replicated from the storage 1 to the storage 2, the probability of losing the data is 0.9, and when data is replicated from the storage 2 to the storage 3, the probability of losing the data Is 0.3. However, since the copy from the storage 1 to the storage 1 is meaningless from the viewpoint of data protection, the risk value is not defined (-).

  The disaster risk value management table 111 is created in advance before data replication. For example, it may be created by calculating the damage probability from statistical information on actual earthquake or tsunami damage, or by calculating using an existing earthquake model or tsunami model, or a numerical simulator You may calculate and create using

  Note that the replication source 400 and the replication source 401 may include all of the storages 101 (here, the four storages 101) or only a part (three of the four storages 101).

  FIG. 5 is a diagram illustrating a configuration example of the data loss weight management table 112. The data loss weight management table 112 stores the value of the data loss weight 502 for each combination of the replication source (replication source storage) 500 and the replication destination (replication destination storage) 501.

  The data loss weight 502 is a numerical value indicating the magnitude of data loss damage. The data loss weight 502 is obtained, for example, by multiplying the disaster risk 402 and the data amount of each storage 101.

  In FIG. 5, the data loss weight 502 when the storage 1 with the data amount 1 is the replication source is the same value as the disaster risk 402, and the data loss when the storage 2 with the data amount 2 is the replication source The weight 502 is twice the disaster risk 402.

  The data loss weight table 112 is generated by the data loss weight calculation program 114 prior to the determination of the replication destination. The data loss weight calculation program 114 reads the storage configuration management table 110 and the disaster risk management table 111 and writes the calculated data loss weight 502 as the data loss weight management table 112.

  FIG. 6 is a diagram illustrating a configuration example of the replication configuration management table 113. The replication configuration management table 113 includes a replication source 600 (replication source storage), a first replication destination (first replication destination storage) 601 indicating the first replication destination storage 101 of data stored in the replication source 600, And a second replication destination (second replication destination storage) 602 indicating the second replication destination storage 101 of the data.

  After the computer system 100 determines the replication destination, the management console 103 can display / modify the contents of the replication configuration management table 113. For example, the administrator operates the management console 103 to check whether there is a problem in the duplication destination or customize it according to special requirements.

  6 shows the ID 300 of the storage 101 as the replication source 600, the first replication destination 602, and the second replication destination 603, but the IP address 303 of the storage 101 may be used.

  The replication destination determination program 115 executes a replication destination determination process for determining a first replication destination and a replication destination determination process for determining a second replication destination before executing data replication. As will be described later, although the replication destination to be determined is different, the replication destination determination processing is the same processing.

  FIG. 7 is a flowchart of a replication destination determination process for determining the first replication destination. When the replication destination determination process for determining the first replication destination is started, the replication destination determination program 115 reads the data loss weight management table 112 (S701), executes a combination process to be described later, and stores the storage serving as the first replication destination. The storage ID determined as the first replication destination is written in the first replication destination 601 of the replication configuration management table 113 (S703).

  FIG. 8 is a diagram illustrating a configuration example of a combination work table used when determining a combination of a replication source and a replication destination. The combination work table 800 has a data loss weight evaluation column 801 and a provisional determination column 802 that describes the progress of determination of a combination of a replication source and a replication destination.

  The data loss weight evaluation column 801 is a copy of the data loss weight management table 112. That is, the data loss weight evaluation column 801 is the ID of each storage that is a replication source, the ID of each storage that is a replication destination, and the value of the data loss weight of each combination thereof.

  The provisional determination column 802 is a value for each step for determining a replication destination and a data loss weight evaluated at that step. Each step for determining a replication destination will be described with reference to FIG.

  FIG. 9 is a flowchart of a combination process for determining a combination of a replication source and a replication destination. In the combination processing of the present embodiment, the combination of storages at risk of data loss (high risk) is made as safe as possible.

  When the combination process is started, the replication destination determination program 115 generates a combination work table 800 (S901). In S901, the copy destination determination program 115 copies the data loss weight management table 112 to the data loss weight evaluation column 801. Then, the replication destination determination program 115 creates columns (steps 1 to 4 in this case) indicating steps in the temporary determination column 802 by the same number as the number of replication sources. However, at this time, the data loss weight is not written in the provisional decision sequence 802.

  Next, the copy destination determination program 115 checks the free capacity of each storage 101 that is a copy destination candidate (S902). In S902, the replication destination determination program 115 reads the data amount 301 and the free capacity 302 of the storage configuration management table 110, and the provisional determination column 802 of the combination work table 800, and the free space decreases by the replication data based on the determined combination. After the capacity 302 is calculated, combinations in which the data amount 301 is larger than the free capacity 302 in the data loss weight evaluation column 801 are excluded from the subsequent combination processing.

  The replication destination determination program 115 writes the replication destination with the smallest data loss weight for each replication source in the provisional determination sequence 802 (S903). In the loop processing of S902 to S906, when S903 is executed for the first time (first loop), the replication destination determination program 115 uses the data loss weight 0. 0 in the data loss weight evaluation column 801 when the storage 1 is the replication source. The smallest 0.3 from 9, 0.3, 0.6 is written in the first step sequence of the provisional decision sequence 802. Next, when the storage 2 is a replication source, the smallest 0.6 of the data loss weights 1.8, 0.6, and 0.8 of the data loss weight evaluation column 801 is set to the first of the temporary determination column 802. Write to one step sequence. Similarly, 0.3 is written to the first step sequence for the storage 3 and 0.4 for the storage 4. Thereafter, when S903 is called a plurality of times, the column number is incremented and written to the second step sequence, the third step sequence and the incremental step each time. When there are a plurality of replication destinations having the same data loss weight for each replication source, the average value of each column in the data loss weight evaluation column 801 corresponding to them is calculated, and the data loss weight of the column with the largest average value is calculated. Is written in the first step sequence.

  The replication destination determination program 115 searches the temporary determination sequence 802 for the replication source with the largest data loss weight, and determines the replication destination for the replication source (S904). In the loop processing of S902 to S906, when executing S904 for the first time, the replication destination determination program 115 includes the data loss weights 0.3, 0.6, 0.3, and 0.4 in the first step sequence. The combination corresponding to the largest 0.6, in which the storage 2 is the replication source and the replication destination is the storage 3, is determined. When there are a plurality of combinations of the replication source and the replication destination having the same largest data loss weight in the provisional decision column 802, the average value of each row in the data loss weight evaluation column 801 is calculated, and the column having the largest average value is calculated. It is assumed that the combination having the data loss weight is determined.

  The replication destination determination program 115 excludes the determined replication source from the combination work table (S905). When S904 is first executed, the replication destination determination program 115 excludes the rows corresponding to the case where the storage 2 is the replication source in the data loss weight evaluation column 801 and the provisional determination column 802 from the subsequent combination processing targets.

  The replication destination determination program 115 checks whether replication destinations have been determined for all replication sources (S906). If it is determined, the copy destination determination program 115 proceeds to S907. If it has not been determined, the replication destination determination program 115 proceeds to S902. When S904 is executed for the first time, the replication destination determination program 115 determines a combination with the storage 2 as the replication source and the storage 3 as the replication destination, but the combination with the replication destinations with the other three storages as the replication sources. Since it has not been determined, the process proceeds to S902. Finally, when S904 is executed, the replication destination determination program 115 sets the storage 1 and storage 4, the storage 2 and storage 3, the storage 3 and storage 1, and the storage 4 and storage 2 as combinations of the replication source and the replication destination. Determine the combination.

  The replication destination determination program 115 executes a combination replacement process (S907) and ends the process.

  FIG. 10 is a diagram illustrating a configuration example of a replacement work table used by a combination replacement process for switching to a better combination based on the temporarily determined combination. The replacement work table 1000 includes a name column 1001, a combination column 1002, an executable column 1003, and a total loss weight column 1004.

  The name column 1001 indicates the name of a temporary combination created by the replacement work. A combination column 1002 indicates the contents of the combination of the storage duplication source and the duplication destination created by the replacement work. In the figure, for example, “1 → 4” means a combination in which the replication source is the storage 1 and the replication destination is the storage 4.

  The executable column 1003 indicates whether a combination of a replication source and a replication destination is allowed (executable). If the executable column is OK, it indicates that the combination is allowed, and if it is NG, it indicates that the combination is not allowed.

  A total loss weight 1004 indicates a data loss weight for a combination of a replication source and a replication destination determined to be executable. In the example of FIG. 10, the data loss weight of the combination whose name is “before replacement” (ie, “1 → 4”, “2 → 3”, “3 → 1”, “4 → 2”) is as shown in FIG. Respectively, 0.6, 0.6, 0.3 and 0.4. Therefore, the total loss weight column is 1.9 which is the sum of these.

  FIG. 11 is a flowchart of the combination replacement process (S907) in which the combination is replaced with a better combination based on the temporarily determined combination of the replication source and the replication destination. When the combination replacement process is started, the replication destination determination program 115 creates the replacement work table 1000 (S1101). In S1101, the copy destination determination program 115 stores the combination of the name 1001 “before replacement”, the combination 1002 temporarily determined in the memory 107, the executable 1003 “OK”, and the total loss weight “total”. Write the total data loss weights respectively. As shown in FIG. 10, “1 → 4”, “2 → 3”, “3 → 1”, and “4 → 2” are written and executed as the combination 1002 corresponding to the “before replacement” of the name 1001. “OK” is written in 1003, and “1.9” is written in the total loss weight 1004.

  The replication destination determination program 115 generates a combination replacement candidate based on the combination with the largest data loss weight (S1102). In S1102, the replication destination determination program 115 searches for a combination having the maximum data loss weight among the provisionally determined combinations. Next, candidates are generated by replacing the duplication destination in the combination with the largest data loss weight and the duplication destinations for all other duplication sources, and are written in the combination column 1002. In the example illustrated in FIG. 10, with reference to the provisional decision column 802 in FIG. 8, “1 → 4” and “2 → 3” having the largest data loss weight 0.6 among the “before replacement” combination 1002. The two combinations of are retrieved. Then, the combination of 4 that is the replication destination of “1 → 4” and 3 that is the replication destination of “2 → 3” is replaced with 1 that is the replication destination of “candidate 1” and “3 → 1”. A combination obtained by replacing the combination “2”, which is the copy destination of “candidate 2” and “4 → 2”, is set as “candidate 3” and is written in the combination column 1002. Further, for the searched combination “2 → 3”, similarly, the combination obtained by replacing the replication destination 3 with the replication destination 1 of “3 → 1” is “candidate 4” and “4 → 2”. The combination replaced with 2 which is the replication destination of “2” is set as “candidate 5” and written in the combination column 1002. Note that the duplication destination obtained by replacing the duplication destination 3 of “2 → 3” with the duplication destination 4 of “1 → 4” is the same combination “2 → 4” as “candidate 1”, but is omitted. You may write this.

  The copy destination determination program 115 determines whether the combination candidate can be executed (S1103). In step S <b> 1103, the feasibility of all the combination candidates written in the combination column 1002 is checked, and the check result is written in the executable column 1003. One of the conditions that can be executed is that the duplication source and the duplication destination are different combinations. A combination of the same copy source and copy destination has a clearly higher risk of data loss than a combination of different copy source and copy destination, so the copy destination determination program 115 determines that this is not executable. In the example illustrated in FIG. 10, candidate 2, candidate 3, candidate 4, and candidate 5 having the same combination of the replication source and the replication destination are determined to be unexecutable by this check, and NG is written in the executable column 1003. In addition, the copy destination determination program 115 makes the combination where the free capacity of the copy destination is insufficient unexecutable as another feasibility check. As another execution feasibility check, when the number of replicas is 2 or more and the combination replacement process is executed for the second time, a combination having the same replication destination as the first replication destination is executed for a certain replication source. It may be impossible.

  The replication destination determination program 115 calculates the total data loss weight of executable combination candidates (S1104). In S1104, the data loss weight management table 112 is referenced to calculate the total data loss weight of executable combinations among all combination candidates written in the combination column 1002, and write the result in the loss weight total column 1004. . In the example illustrated in FIG. 10, 1.8 that is the sum of the data loss weights (0.3, 0.8, 0.3, 0.4) is written to the loss candidate total 1004 for the executable candidate 1. .

  The copy destination determination program 115 determines whether or not the data loss weight has decreased for the executable combination candidates (S1105). In S1105, the replication destination determination program 115 compares the combination before replacement with the total data loss weight of executable combination candidates, and determines whether the data loss weight has decreased. If there is even one candidate that has decreased, the copy destination determination program 115 proceeds to S1106. Otherwise, the process proceeds to S1107. In the example illustrated in FIG. 10, the combination of candidate 1 has the data loss weight reduced by 0.1 compared to the combination before replacement, and thus the process proceeds to S1106.

  The replication destination determination program 115 determines a combination candidate having the smallest data loss weight as a result of the combination replacement process (S1106), and ends the process. In the example illustrated in FIG. 10, the candidate 1 combination is determined as a result of the combination replacement process.

  When the processing proceeds to S1107, the replication destination determination program 115 determines the combination before replacement as a result of the combination replacement processing (S1107), and ends the processing.

  FIG. 12 is a flowchart of a replication destination determination process for determining the second replication destination 602 of the replication configuration management table 113. When the replication destination determination process for determining the second replication destination is started, the replication destination determination program 115 executes an update process of the data loss weight management table 112 (S1201), and reads the updated data loss weight management table 112 (S1202). ). Next, combination processing to be described later is executed to determine a storage serving as a second replication destination (S1203). Finally, the storage ID determined as the second replication destination is written in the second replication destination 602 of the replication configuration management table 113 (S1204), and the replication destination determination process for determining the second replication destination is terminated.

  FIG. 13 is a diagram illustrating an example of the data loss weight management table 1300 obtained by updating the data loss weight of the data loss weight management table 112 in the present embodiment. The difference between the data loss weight management table 1300 after the update and the data loss weight management table 112 before the update is the value of the data loss weight. Of the data loss weights 1303, the combination that is “determined” has already been determined by the replication destination determination process for determining the first replication destination (candidate 1 in FIG. 10). It means to exclude from the target of the combination to determine. The data loss weight 1303 of other combinations is a numerical value indicating the magnitude of data loss damage when the second copy destination is determined after copying to the first copy destination.

  For example, the data loss weight 1303 when the storage 1 in FIG. 13 is the replication source is the data loss with the storage 3 determined as the first replication destination in the data loss weight 502 before update when the storage 1 is the replication source. Obtained by multiplying the loss risk by 0.3. For example, the data loss weight value 1303 of the combination where the storage 1 is the replication source and the storage 2 is the second replication destination is the value 0 of the data loss risk with the storage 3 to the value 0.9 of the data loss weight 502 before the update. .3 multiplied by .3.

  FIG. 14 is a flowchart of data loss weight update processing for updating the data loss weight. When the data loss weight update process is started, the replication destination determination program 115 reads the data loss weight management table 112 and the replication configuration management table 113 (S1401).

  The replication destination determination program 115 reads the first replication destination of each replication source from the replication configuration management table 113 and sets the combination with the first replication destination of the data loss weight management table 112 to the next update weight value (S1403). Excluded from the calculation target (S1402).

  The copy destination determination program 115 calculates an update weight value (S1403). As described in the explanation of FIG. 13, the update weight value is obtained by multiplying the data loss weight recorded in the data loss management table 112 before update by the disaster risk 402.

  The replication destination determination program 115 writes the calculated update weight value in the updated data loss risk management table 1300 (S1404), and ends the process.

  By using the data loss weight management table 112 (1300 in FIG. 13) in which the data loss weight has been updated, the replication destination determination program in FIG. 7 is executed again to obtain a combination of the replication source and the second replication destination storage. The storage of the second replication destination 602 corresponding to each storage of the replication source 600 is determined. The above is the replication destination determination process of the second replication destination in the present embodiment by the replication destination determination program 115.

  The replication control program 116 refers to the replication configuration management table 113 and instructs the replication program 205 of each storage to replicate. As a result, the replication program 205 of each storage of the replication source 600 replicates data to the first replication destination 601 and the second replication destination 602.

  In this embodiment, the case where the number of replicas is 2 (redundancy is 3) has been described as an example. However, the same method can be used when the number of replicas is 3 or more (redundancy is 4 or more). For example, when the number of replicas is 3, as a procedure for determining the third replication destination, a procedure for simply determining the second replication destination may be additionally executed.

  In this embodiment, the data loss weight 502 is obtained by multiplying the disaster risk 402 by the data amount of each storage 101, but it may be simply the disaster risk 402. The data loss weight 502 may be a data safety weight (that is, a weight indicating that the larger the value is, the safer the data is).

  FIG. 15 is a diagram showing the effect of this embodiment. The data loss ratio graph in FIG. 15A reflects the determination of the replication destination and batch processing (for example, the data loss weight updated with the determination of the first replication destination) in the determination of the second replication destination by repeating this embodiment. This is experimental data showing how the data loss rate increases with respect to the increase in the damaged storage rate when each of the above is used. In the experiment, a computer system 100 with 135 storages 101 was used. In the computer system 100, after duplicating data and replicating data, the damage of storage was simulated using an earthquake simulator, and the data loss rate was calculated by counting the data loss and the remaining from the result. The data loss ratio graph in FIG. 15A shows that the data loss ratio between the iterative processing and the batch processing in this embodiment is almost the same regardless of the damaged storage ratio. That is, it can be seen that the determination of the replication destination by the iterative processing of this embodiment has the same data protection performance as the determination of the replication destination by the batch processing.

  The time graph for determining the replication destination in FIG. 15B is used to determine the replication destination with respect to the increase in the number of storages 101 constituting the computer system 100 when the iterative processing and batch processing of this embodiment are used. This is experimental data showing how such time increases. The time graph for determining the replication destination in FIG. 15B shows that the time required for determining the replication destination by the iterative processing of this embodiment is smaller than that by the batch processing, regardless of the number of storages, and is several thousand minutes. It shows that there are even cases where it is less than one. That is, according to the iterative process of the present embodiment, it can be seen that the duplication destination can be determined in a shorter time than the batch process.

  According to the present embodiment, the data replication destination storage can be determined in a short time. Therefore, when operating a computer system (large-scale distributed system) in which many storages are geographically distributed, replication of each storage is performed. First, it is possible to determine a high data protection performance in a short time. Therefore, the user can quickly change the duplication destination each time the requirements of the large-scale distributed system change. At that time, the data protection effect is kept almost the same as when batch processing is used.

  In this embodiment, a description will be given of repetitive determination of a data replication destination having a different redundancy for each storage. In the present embodiment, a description will be given centering on differences from the first embodiment.

  Unlike the first embodiment, this embodiment replicates data with different data redundancy for each copy source storage. In this embodiment, the computer system 100 determines the data loss weight in advance in the process of repeatedly determining the data replication destination for each storage serving as the replication source at each stage of determining the replication source described in the first embodiment. If the weight is equal to or less than the given allowable data loss weight, there is provided a configuration in which no more data replication destinations are provided and replication is performed with different redundancy for each storage serving as a replication source. In other words, when the data loss weight in the combination of the copy source storage and the nth copy destination storage exceeds the predetermined allowable data loss weight (when the predetermined allowable data loss weight is not satisfied), the data loss The weight is updated, and the (n + 1) th replication destination storage is determined. As a result, the configuration is such that replication is performed with different redundancy for each storage serving as a replication source.

  FIG. 16 is a diagram illustrating a configuration example of a data redundancy management screen displayed on the management console 103 in the present embodiment. The data redundancy management screen 1600 is a fixed redundancy radio button 1605 for setting data replication with the same redundancy for all storages 101 of the computer system 100. Data with different redundancy for each storage A variable redundancy radio button 1601 for setting for duplicating is displayed. Further, the data redundancy management screen 1600 includes a maximum redundancy text box 1602 for setting the maximum redundancy allowed in response to selection of the variable redundancy radio button 1601, an allowable data loss weight text box 1603, and a management server. An OK button 1604 for notifying the completion of setting is provided to 102. It should be noted that the data loss weight exceeding the value input in the allowable data loss weight text box 1603 is not allowed.

  An operator such as a system administrator operates the management console 103 to display the data redundancy management screen 1600 and selects the fixed redundancy radio button 1605 or the variable redundancy radio button 1601. When the variable redundancy radio button 1601 is selected, after setting the allowable maximum redundancy 1602 and the allowable data loss weight 1603, pressing the OK button 1604 completes the setting. The management console 103 stores the set contents on the memory 107 of the management server 102. In this embodiment, the variable redundancy radio button 1601 is selected.

  FIG. 17 is a diagram illustrating an example of a data loss weight management table obtained by updating the data loss weight in the present embodiment. A difference between the data loss weight management table 1700 obtained by updating the data loss weight and the data loss weight management table 1300 in the first embodiment is that “exclusion” can be stored in the data loss weight value 1703. The combination of storage of the replication source 1701 and the replication destination 1702 in which the data loss weight 1703 is “excluded” is the data loss allowed by the storage data loss weight value determined by the replication destination determination program 115 as the first replication destination. This means that the copy destination determination program 115 does not determine the second copy destination (excludes processing) because it is equal to or less than the value input to the weight 1603.

  In the example shown in FIG. 16, the allowable data loss weight 1603 is 0.3. Since the data loss weight value between the storage 1 and the storage 3 that is the first replication destination is 0.3 (see FIG. 5), which is less than or equal to the allowable data loss weight 1603, the storage 1 Excluded from the target to determine the replication destination. On the other hand, the data loss weight value between the storage 2 and the storage 4 that is the first replication destination is 0.8 (see FIG. 5), which is larger than the allowable data loss weight 1603 value. It is included in the target for determining the second replication destination. Similarly, storage 3 is excluded from its target and storage 4 is included in its target.

  FIG. 18 is a flowchart showing the procedure for updating the data loss weight in this embodiment. The difference from the first embodiment is that in FIG. 18, S1803 is added between S1402 and S1403 in FIG.

  When the data loss weight update process is started, the replication destination determination program 115 reads the data loss weight management table 112 and the replication configuration management table 113 (S1801). The replication destination determination program 115 reads the first replication destination of each replication source from the replication configuration management table 113, and selects the combination with the first replication destination from the data loss weight management table 112 to calculate the update weight value in S1804. (S1802).

  The replication destination determination program 115 reads the value of the data loss weight allowed from the memory 107 (the value input to the allowable data loss weight 1603 and stored in the memory 107 described above) and described in the explanation of FIG. As described above, the data loss weight in the combination with the first replication destination of each replication source is compared with this. As a result, the copy source having the combination of the allowable data loss weight 1603 or less is excluded from the calculation target of the S1804 update weight value (S1803).

  The replication destination determination program 115 calculates an update weight value (S1804), writes the calculated update weight value as the updated data loss weight management table 1700 (S1805), and ends the process. The above is the copy destination determination process in the present embodiment by the copy destination determination program 115.

  In this example, the method of replicating data with different redundancy for each storage is shown. However, the same method as in this example is used so that the data (ie, file or directory) has different redundancy. May be.

  According to the present embodiment, the copy source determined to be safe based on the allowable data loss weight value does not create any more copies, so the data can be efficiently copied to the safe storage, and the copy data is stored. Storage capacity can be saved.

  100: Computer system, 101: Storage, 102: Management server, 103: Management console, 110: Storage configuration management table, 111: Disaster risk management table, 112: Data loss weight management table, 113: Replication configuration management table, 114: Data loss weight calculation program, 115: replication destination determination program, 116: replication control program.

Claims (10)

Corresponding to each of the plurality of storages and the plurality of storages that are connected to the plurality of storages via a network and storing the data, the data storage weight of the data is used, One of the plurality of storages for storing data replication data is determined as a first replication destination storage, the data loss weight is updated using a given disaster risk, and the updated data loss weight is set. And a management server that determines the other one of the plurality of storages as the second replication destination storage that stores the replication data in correspondence with the replication source storage that stores the data. And
The management server tentatively determines a combination of the replication source storage and the first replication destination storage, and tentatively determines the replication source storage for the combination with the largest data loss weight among the tentatively determined combinations. The combination of the first replication destination storage temporarily determined and the first replication destination storage is replaced, and the first replication destination storage temporarily determined from the replaced combination is replaced with the first replication destination storage using the data loss weight of the data. A computer system characterized by determining as . The management server updates the data loss weight using a given disaster risk when the data loss weight does not satisfy a predetermined allowable data loss weight, and uses the updated data loss weight, 2. The second replication destination storage is determined as another second storage among the plurality of storages storing the replication data corresponding to the replication source storage storing data. The computer system described in 1. 2. The computer according to claim 1 , wherein when updating the data loss weight, the management server excludes an update of a data loss weight corresponding to a combination of the replication source storage and the first replication destination storage. system. The computer system according to claim 3 , wherein the management server determines the second replication destination storage by excluding the first replication destination storage. The data replication data using the data loss weight of the data corresponding to the replication source storage of each of the plurality of storages storing data to a management server connected to the plurality of storages via a network One of the plurality of storages is stored as a first replication destination storage, the data loss weight is updated using a given disaster risk, and the updated data loss weight is used to Corresponding to the replication source storage storing the data, the procedure for determining the other one of the plurality of storages for storing the replication data as a second replication destination storage is executed .
The management server temporarily determines a combination of the replication source storage and the first replication destination storage, and temporarily determines the replication source storage for the combination with the largest data loss weight among the provisionally determined combinations. The combination of the first replication destination storage temporarily determined and the first replication destination storage is replaced, and the first replication destination storage temporarily determined from the replaced combination is replaced with the first replication destination storage using the data loss weight of the data. A program for executing the procedure to determine as . In the management server, if the data loss weight does not satisfy a predetermined allowable data loss weight, the data loss weight is updated using a given disaster risk, and the updated data loss weight is used to update the data loss weight. The claim for executing the procedure which determines the other one of the plurality of storages storing the replication data as the second replication destination storage corresponding to the replication source storage storing the data. 5. The program according to 5 . To the management server, when said updating data loss weight, according to claim 6 for executing exclude steps to update the data loss weight corresponding to the combination of the said replication source storage first destination storage program. The program according to claim 7 , wherein the management server is caused to execute a procedure for determining the second replication destination storage by excluding the first replication destination storage. In a computer system having a plurality of storages and a management server connected to the plurality of storages via a network,
The management server is
Data replication data using the data loss weight of the data calculated for each combination of the replication source storage and the replication destination storage corresponding to the replication source storage of each of the plurality of storages storing data And determining one of the plurality of storages as a first replication destination storage,
The data loss weight calculated for each combination of the replication source storage and the replication destination storage is updated using the determined first replication destination storage, and the data is stored using the updated data loss weight A computer system, wherein another one of the plurality of storages storing the duplicate data is determined as a second duplication destination storage corresponding to the duplication source storage. To a management server connected to multiple storages via a network,
Data replication data using the data loss weight of the data calculated for each combination of the replication source storage and the replication destination storage corresponding to the replication source storage of each of the plurality of storages storing data And executing a procedure for determining one of the plurality of storages as a first replication destination storage,
The data loss weight calculated for each combination of the replication source storage and the replication destination storage is updated using the determined first replication destination storage, and the data is stored using the updated data loss weight A program for executing a procedure of determining another one of the plurality of storages that stores the replication data as a second replication destination storage corresponding to the replication source storage.

Images