JP2010205100A - Management server, backup system, backup method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a backup system which optimizes the amount of difference data while reducing traffic between a management server and a management target machine. <P>SOLUTION: A management server 1000 includes: a full backup data storage means 1110 wherein full backup data are stored; a working area 1300 wherein hash data including a block size by which the full backup data are divided into a plurality of blocks, and a plurality of hash values corresponding to respective blocks is stored; a boot image 1220 by which the management target machine extracts difference data between full backup data and present data by using hash data; a data transmission/reception means 1500 which transmits hash data to the management target machine after the management target machine is invoked using the boot image, and receives difference data from the management target machine; and a dividing/combining means 1400 which uses full backup data and difference data to calculate a block size for next backup. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a method for backing up data stored in a managed machine, and more particularly, to a differential backup method in which differential data changed from previously backed up data is a backup target.

A server for operating a business service or a business terminal (personal computer) is usually operated to back up all important data or disks in order to prepare for an unexpected accident. In particular, with respect to a server operating a business service, operation that suppresses the server stop as much as possible is required due to the recent non-stop service flow. Therefore, even when stopping, it is required to shorten the time as much as possible.
For example, when backing up only important data, operations such as stopping business applications and securing application quiesce points (state that is semantically consistent) and then backing up application data Has been made. Also, when backing up a system (OS: Operating System), the system is stopped and backed up.

Under such circumstances, various differential backup methods have been devised in order to minimize the downtime of business applications and systems. For example, Patent Document 1 describes a technique for copying only difference data of a business volume to a copy volume.
In the method of Patent Document 1, an update flag is prepared for each block of a transaction volume (copy source), and the flag is managed at the update timing. Therefore, the previous backup was executed (the volume is disconnected). The block updated from the time point can be grasped. For this reason, it is possible to specify which block should be copied when the next backup is executed.

JP 2000-330730 A

  However, according to the method disclosed in Patent Document 1, a difference calculation mechanism (storage device, file system) is required on the backup target. For this reason, the cost of system construction has become high. On the other hand, in general-purpose disk backup methods that do not require a special mechanism for backup, a full backup has always been required. For this reason, the backup time and the amount of backup data have been enlarged.

When attempting to implement differential backup for these, a method of calculating differential data on the management server after collecting a full backup, or sending the full backup data collected last time to the managed machine I had to take the method of calculating the difference above. However, these methods are not practical because the amount of communication data increases.
When calculating the difference at the disk image level, the appropriate value for the block size to be calculated depends on the file system mounted on the disk or the business application running on the file system. The amount of difference data was not the optimum amount.

  An object of the present invention relates to a backup that optimizes the amount of differential data while suppressing the amount of communication between a management server and a managed machine for a managed machine that does not have a difference calculation mechanism (storage device, file system). Is to provide technology.

  One aspect of the management server according to the present invention is a management server that backs up data of a managed machine, full backup data storage means for storing the full backup data of the managed machine, and a plurality of the full backup data. Hash data storage means for storing hash data including a block size to be divided into blocks and a plurality of hash values corresponding to each block, and difference data between the full backup data and the current data, using the hash data Boot image storage means for storing a boot image to be extracted by the managed machine, transmission means for sending the boot image to the managed machine when the managed machine starts backup, and the boot image To start the managed machine After that, the hash data is transmitted to the managed machine, the data transmission / reception means for receiving the differential data extracted by the managed machine, and the full backup data and the differential data are used for the next backup. Calculates the block size to be used, stores the latest full backup data reflecting the difference data in the full backup data storage means, and supports each block obtained by dividing the latest full backup data using the calculated block size Dividing / combining means for calculating a hash value to be stored and storing the calculated block size and the calculated hash value in the hash data storage means.

  One aspect of the backup method according to the present invention is a backup method in which data of a managed machine is backed up by a management server. The management server includes the above-described units, and the managed machine includes the boot image and the boot image. Data transmission / reception means for receiving the hash data and transmitting the difference data, and difference data creation means for calculating hash data of backup target data and creating difference data by comparing with the received hash data.

  One aspect of the backup method according to the present invention is a backup method for backing up data of a managed machine, wherein the full backup data of the managed machine is backed up to a full backup data storage means, and the full backup data is stored in a plurality of ways. Hash data including a block size to be divided into blocks and a plurality of hash values corresponding to each block is stored in a hash data storage unit, and difference data between the full backup data and current data is used as the hash data. The boot image to be extracted by the managed machine is stored in boot image storage means, and when the managed machine starts backup, the managed machine is restarted using the boot image, and the managed machine To the hashday The difference data is extracted, the block size used for the next backup is calculated using the full backup data and the difference data, and the latest full backup data reflecting the difference data is calculated as the full data. A hash value corresponding to each block obtained by dividing the latest full backup data using the calculated block size is stored in the backup data storage means, and the calculated block size and the calculated hash value are stored in the hash data Store in the means.

  One aspect of the program according to the present invention is a program that implements a backup for backing up data of a managed machine, wherein the computer backs up the full backup data of the managed machine to a full backup data storage unit; A process of storing hash data including a block size for dividing the full backup data into a plurality of blocks and a plurality of hash values corresponding to each block in a hash data storage unit; and when the managed machine starts backup A process of transmitting a boot image for extracting the difference data between the full backup data and the current data to the managed machine using the hash data, and restarting the managed machine; and the managed machine To the hash The difference data is extracted, the process of calculating the block size used for the next backup using the full backup data and the difference data, and the latest full backup data reflecting the difference data , The hash value corresponding to each block obtained by dividing the latest full backup data using the calculated block size, the calculated block size and the calculated hash Storing the value in the hash data storage means.

  The present invention relates to a backup that optimizes the amount of difference data while suppressing the amount of communication between the management server and the managed machine for a managed machine that does not have a difference calculation mechanism (storage device, file system). Technology can be provided.

It is a figure showing typically the system configuration concerning the present invention. FIG. 3 is a block diagram schematically illustrating a logical structure of a program that operates in the management server according to the first embodiment. It is a figure explaining the image of the full back data of a disk area | region. It is a figure explaining difference data. It is a figure showing the data structure of the difference data stored in a difference data storage means. It is a figure which shows the specific example of a hash value. It is a figure showing the data structure of the hash data stored in a working area. It is a figure showing the logic which determines the block size which divides | segments backup data. It is a figure showing an example of the change of the amount of difference data at the time of assuming that the difference data collected this time was divided | segmented by the block size of 1/2 times. It is a figure showing an example of the change of the amount of difference data at the time of assuming that the difference data collected this time was divided | segmented by 2 times the block size. FIG. 3 is a block diagram schematically illustrating a logical structure of a boot image that operates on a managed machine. It is a flowchart which shows the operation example of the whole process of a management server and a management object machine. It is a flowchart which shows the operation example of the differential backup process of a management object machine. It is a flowchart which shows the operation example of the division | segmentation / coupling | bonding process of a management server. It is a flowchart which shows the operation example of the division | segmentation efficiency calculation process of a management server. It is a flowchart which shows the operation example of the coupling efficiency calculation process of a management server. It is a flowchart which shows the operation example of the hash data calculation process of a management server. It is a flowchart which shows the operation example of the hash data comparison process of a management server. 2 is a diagram illustrating data showing a specific example of the first embodiment, where (a) shows an example of full back data stored in the full backup data storage means, and (b) shows an example of hash data stored in the working area. FIG. . An example of data when the managed machine creates difference data is shown, (a) is an example of hash data calculated by the hash data calculation means, and (b) is an example of difference data created by the difference data creation means. An example of data when the management server creates difference data is shown, (a) is an example of hash data of the previous area, and (b) is an example of hash data in the difference data. An example of data when the management server performs the coupling efficiency calculation is shown. (A) is a block corresponding to differential data in the bull backup data, and (b) shows an example of differential data.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. For clarity of explanation, the following description and drawings are omitted and simplified as appropriate. In the drawings, components having the same configuration or function and corresponding parts are denoted by the same reference numerals and description thereof is omitted.

  The present invention provides a backup method applicable when the managed machine does not have a difference calculation mechanism (hardware or software). Further, a method of backing up differential data different from the data backed up last time is used. Therefore, the management server transmits the hash data of the full backup data collected last time to the management target machine and calculates the difference data on the management target machine when executing the differential backup. Further, the management server calculates an appropriate value of the block size for the difference calculation after executing the differential backup, and automatically changes the block size if it is not an appropriate value. Embodiments will be described below with reference to the drawings.

(Embodiment 1)
FIG. 1 schematically shows the system configuration of this embodiment, which includes a management server 1000, a management network 2000, and managed machines 3000 to 5000.
The management server 1000 is hardware equipped with software for managing and controlling the managed machines 3000 to 5000 via the management network 2000.
The management network 2000 is a network infrastructure through which the management server 1000 manages and controls the managed machines 3000 to 5000.
The managed machines 3000 to 5000 are hardware managed and controlled by the management server 1000.

  FIG. 1 is an illustration for schematically explaining the present embodiment, and does not limit the system configuration. In other words, the managed machine can be configured as an arbitrary number according to the system to be managed / controlled.

FIG. 2 schematically shows the logical structure of the program of this embodiment that operates in the management server 1000. The structure of the management server 1000 includes a backup data storage area 1100, a boot image storage area 1200, a working area (hash data storage means) 1300, a division / combination means 1400, a data transmission / reception means 1500, and a boot image transmission means 1600.
Here, the backup data storage area 1100 and the boot image storage area 1200 correspond to a nonvolatile storage device such as a magnetic disk and a DBMS (DataBase Management System) that manages data stored therein.

The working area 1300 corresponds to a volatile storage area such as a main memory.
The dividing / combining unit 1400, the data transmitting / receiving unit 1500, and the boot image transmitting unit 1600 correspond to program modules executed by a CPU (Central Processing Unit) or the like. The program module is loaded into the memory, and a group of instructions is executed under the control of a CPU (Central Processing Unit) to execute the processing of each means.

The backup data storage area 1100 includes full backup data storage means 1110 and differential data storage means 1120.
The boot image storage area 1200 includes a boot image storage unit 1210, and the boot image 1220 is stored in the boot image storage unit 1210.
The dividing / combining unit 1400 includes a dividing efficiency calculating unit 1410, a hash data calculating unit 1420, a hash data comparing unit 1430, and a combining efficiency calculating unit 1440.

Hereinafter, each component will be described in detail.
The full backup data storage unit 1110 stores the full backup data of the managed machine.
The full backup data corresponds to a copy of the disk area of the managed machine, and refers to a full copy of data from address 0 to the maximum disk address N as shown in FIG. 3, for example.

The difference data storage unit 1120 stores difference data between the full backup data collected this time from the managed machine and the full backup data collected last time. The difference data is data that is obtained by dividing the full-back data collected this time and the full-back data collected last time into predetermined block sizes, comparing the data stored in the same address block, and collecting different blocks. is there.
For example, as shown in Fig. 4, if there is full backup data of size N, the previous collection and the current collection are divided by the same block size (n), and the data in the block exactly matches the previous collection In this case, “differential data” is defined as data obtained by collecting blocks having “difference” when “no difference” is set, and even if some of the data in the block is different from the previous collection, “difference exists”.
FIG. 5 shows a data structure representing the difference data. The difference data includes a “size” corresponding to the block size to be divided, an “address” corresponding to the head address of the block, and “data” corresponding to the actual block data.

  The boot image 1220 stored in the boot image storage unit 1210 corresponds to a boot image (differential backup unit) transmitted to the managed machine in order to execute the differential backup. When the management server 1000 transmits the boot image 1220 to the managed machine and restarts the managed machine, various programs in the boot image 1220 operate and differential backup is executed. Therefore, it can be said that the boot image 1220 is a program that realizes differential backup. Details of the boot image 1220 will be described later with reference to FIG.

In the working area 1300, hash data corresponding to the full backup data stored in the full backup data storage unit 1110 is stored as advance preparation for differential backup.
Here, the hash data is defined as “hash data” that is obtained by dividing the hash target data by a certain block size (n) and collecting the hash values of each block. FIG. 6 shows a specific example of the hash value.
FIG. 7 shows a data structure representing hash data. The hash data is composed of a table storing the “size” corresponding to the block size to be divided, the head address of the block, and the hash value corresponding to the block.

  The dividing / combining means 1400 is a program module that calculates the amount of difference data when the block size to be divided is increased or decreased with reference to the backup data storage area 1100 and determines the optimum block size at the next difference backup. is there.

  FIG. 8 shows the logic for determining the optimal block size. Each line indicates whether the difference data amount decreases, is the same, or increases when the current block size is divided. In addition, each column indicates whether the difference data amount decreases, is the same, or increases when the current block sizes are combined. If the difference data amount decreases when the current block size is divided and the difference data is the same or increased when combined, the block size is divided. If the difference data is the same when the current block size is divided and combined, the block sizes are combined. If the difference data amount is the same when the current block size is divided and the difference data increases when combined, the block size is not changed and the current state is maintained. N / A (Non Applicable) means a combination that cannot be applied.

  In FIG. 8, the difference data amount obtained when the difference data collected this time is assumed to be divided by a half block size (division, FIG. 9) and the case where it is assumed that the difference data is divided by a double block size. The calculation result of the difference data amount (combined, FIG. 10) determines how the divided block size of the next difference data is to be made.

The division efficiency calculation means 1410 calculates whether or not the difference data amount decreases when the division block size is halved.
The hash data calculation unit 1420 creates hash data of the specified calculation target area with the specified block size.
The hash data comparison unit 1430 compares the two specified hash data, and calculates a list of different hash values.
The coupling efficiency calculation unit 1440 calculates whether or not the difference data amount increases when the divided block size is doubled.

  The data transmission / reception unit 1500 is a program module that transmits hash data to a managed machine and receives differential data from the managed machine.

  The boot image transmission unit 1600 is a program module that transmits a boot image 1220 to a managed machine and restarts the managed machine with the boot image. For example, it may be a server program according to the PXE (Preboot eXecution Environment) protocol specified in RFC (Request For Comments) 4578, or a boot image is sent to a so-called client program operating on a managed machine by the Push method. There may be.

  Since a method of transmitting the boot image 1220 to the managed machine and restarting the managed machine with the boot image can be sufficiently performed by a known technique, in this embodiment, the boot image is transmitted and the boot image is transmitted. The method for restarting managed machines is not specified.

  FIG. 11 schematically illustrates the logical structure of the boot image 1220 that operates on the managed machine. The boot image 1220 includes data transmission / reception means 1221, difference data creation means 1222, hash data calculation means 1223, and hash data comparison means 1224.

  The data transmission / reception unit 1221, the difference data creation unit 1222, the hash data calculation unit 1223, and the hash data comparison unit 1224 correspond to program modules executed by the CPU or the like.

The data transmission / reception unit 1221 receives hash data corresponding to the full backup data collected last time from the management server 1000 and transmits the difference data created on the managed machine to the management server 1000.
The difference data creation unit 1222 performs a process of creating difference data from the hash data calculated from the disk area of the managed machine and the hash data corresponding to the full backup data for the previous collection received from the management server 1000.
The hash data calculation unit 1223 creates hash data of the disk area of the managed machine with the block size in the hash data received from the management server 1000.
The hash data comparison unit 1224 compares the hash data corresponding to the previously collected full backup data received from the management server 1000 and the hash data calculated from the disk area of the managed machine.

  Next, the operation of this embodiment will be described with reference to FIG. 2 and FIGS. When a differential backup is instructed by a user operation or the like (step S11 in FIG. 12), the boot image transmission unit 1600 in FIG. 2 transmits the boot image 1220 to the managed machine and restarts the managed machine. (Step S12 in FIG. 12). When the start of differential backup processing is confirmed in the managed machine, the data transmission / reception means 1500 in FIG. 2 manages the hash data stored in the working area 1300 (hereinafter, this hash data is referred to as “hash data 1”). Transmit to the target machine (step S13 in FIG. 12).

The managed machine receives the boot image 1220 (step S21 in FIG. 12) and is restarted. The restarted managed machine is started with the boot image 1220 of FIG. 11 (step S22 of FIG. 12), and the differential backup process (FIG. 13) is started (step S23 of FIG. 12).
When the differential backup process is started, the data transmitting / receiving unit 1221 in FIG. 11 receives the hash data 1 of the full backup data collected last time from the management server 1000 (step A1 in FIG. 13).
After receiving the hash data 1 from the management server 1000, the differential data creation unit 1222 in FIG. 11 calculates hash data by inputting “the disk area of the managed machine” as the calculation target area and “the size of the hash data 1” as the size. Means 1223 is called (step A2 in FIG. 13). The specific operation of the hash data calculation means 1223 is as shown in FIG. First, the hash data calculation unit 1223 accepts a calculation target area (here, “disk area of the management target machine”) and a size (here, “size of hash data 1”), and sets the calculation target area as a block having a size. (Step E1 in FIG. 17). Next, the hash data calculation means 1223 calculates a hash value of each block (step E2 in FIG. 17), and outputs a set of size + hash value as “hash data” (step E3 in FIG. 17).

  After calculating the hash data of the disk area of the managed machine (hereinafter, this hash data is referred to as “hash data 2”), the differential data creation unit 1222 uses the hash data comparison unit 1224 to generate the hash data 1 and the hash data. 2 are compared (step A3 in FIG. 13), blocks having different hash values are collected to create difference data, and the difference data is transmitted to the management server 1000 using the data transmitting / receiving means 1221 (step A4 in FIG. 13). . The specific operation of the hash data comparison unit 1224 is as shown in FIG. First, the hash data comparison unit 1224 accepts two hash data (here, hash data 1 and hash data 2), and compares the hash value of the corresponding hash data 2 for each hash value of the hash data 1 (see FIG. 18 Step F1), a list of items having different hash values is output (Step F2 in FIG. 18).

  When the data transmission / reception unit 1500 receives the difference data from the managed machine, the management server 1000 in FIG. 2 stores the difference data in the difference data storage unit 1120 (step S14 in FIG. 12), and the division / combination processing (FIG. 14). ) Is started (step S15 in FIG. 12).

  2 extracts an area (previous area) corresponding to the differential data stored in the differential data storage means 1120 from the full backup data stored in the full backup data storage means 1110. (Step B1 in FIG. 14 and Step C1 in FIG. 15).

  After extracting the previous area, the division efficiency calculation means 1410 calls the hash data calculation means 1420 with “previous area” as the calculation target area and “1/2 times the size in the difference data” as the input. (Step C2 in FIG. 15), the hash data 3 is created. Thereafter, the division efficiency calculation unit 1410 calls the hash data calculation unit 1420 with “data in the difference data” as the calculation target area and “1/2 times the size in the difference data” as the size (FIG. 15). Step C3), the hash data 4 is created. Specifically, the hash data calculation unit 1420 calculates a calculation target area (here, “previous area” or “data in difference data”) and size (here, “½ times the size in difference data”). And the calculation target area is divided into blocks of size (step E1 in FIG. 17). Next, the hash data calculation unit 1420 calculates a hash value of each block (step E2 in FIG. 17), and outputs a set of size + hash value as “hash data” (step E3 in FIG. 17).

  After creating the hash data 3 and the hash data 4, the division efficiency calculation unit 1410 calls the hash data comparison unit 1430. The hash data comparison unit 1430 in FIG. 2 compares the hash data 3 and the hash data 4 (step C4 in FIG. 15). Specifically, the hash data comparison unit 1430 receives two pieces of hash data (here, hash data 3 and hash data 4), and compares the hash value of the corresponding hash data 4 for each hash value of the hash data 3. (Step F1 in FIG. 18), and a list of items having different hash values is output (Step F2 in FIG. 18).

Next, the division efficiency calculation unit 1410 calculates the difference data amount when the block size is ½ (step C5 in FIG. 15).
When the calculation result is smaller than the original difference data, the division efficiency calculation unit 1410 determines that “the difference data size is reduced” (step C8 in FIG. 15), and otherwise, “the difference data size is not reduced”. Judgment is made (step C7 in FIG. 15).

  As a result of the division efficiency calculation process, when the difference data amount decreases (Y in step B2 in FIG. 14), the division / combination means 1400 determines that the division size is ½ times the proper value, After the difference data is merged with the full backup data in the full backup data storage unit 1110 (step B9 in FIG. 14), hash data to be transmitted to the next managed machine is created (step B10 in FIG. 14). Specifically, the dividing / combining unit 1400 calls the hash data calculating unit 1420 with “full backup image” as the calculation target area and “1/2 times the size” as the input, and sets the next management target machine. Hash data to be transmitted is created (E1 to E3 in FIG. 17). The dividing / combining means 1400 stores the created hash data in the working area 1300 and ends the process.

When the difference data amount does not decrease as a result of the division efficiency calculation process (N in Step B2 in FIG. 14), the coupling efficiency calculation unit 1440 in FIG. 2 converts the full backup data stored in the full backup data storage unit 1110 to The block is divided into blocks having a size twice the size of the difference data stored in the difference data storage means 1120, and a block (combined area) including the data in the difference data is extracted (step B3 in FIG. 14, FIG. 16 step D1).
After extracting the combined area, the amount of data in the combined area and the difference data is compared (step D2 in FIG. 16). If the size is the same, it is determined that “the data size does not increase” (step D5 in FIG. 16). Otherwise, it is determined that “the data size increases” (step D4 in FIG. 16).

  If the difference data amount increases as a result of the coupling efficiency calculation process (Y in step B4 in FIG. 14), the dividing / combining unit 1400 determines that the current division size is an appropriate value, and the difference data is converted to the full backup data. After merging with the full backup data in the storage unit 1110 (step B7 in FIG. 14), hash data to be transmitted to the next managed machine is created (step B8 in FIG. 14). Specifically, the dividing / combining unit 1400 calls the hash data calculating unit 1420 with “full backup image” as the calculation target area and “the same size as the previous time” as the size, and transmits them to the next managed machine. Hash data is created (E1 to E3 in FIG. 17). The dividing / combining means 1400 stores the created hash data in the working area 1300 and ends the process.

  If the difference data amount does not increase as a result of the coupling efficiency calculation process (N in step B4 in FIG. 14), the dividing / combining means 1400 determines that the division size is twice as appropriate, and the difference data Are merged with the full backup data in the full backup data storage means 1110 (step B5 in FIG. 14), and hash data to be transmitted to the next managed machine is created (step B6 in FIG. 14). Specifically, the dividing / combining unit 1400 calls the hash data calculating unit 1420 with “full backup image” as the calculation target area and “double size” as the size as input, and transmits to the next managed machine. Hash data is created (E1 to E3 in FIG. 17). The dividing / combining means 1400 stores the created hash data in the working area 1300 and ends the process.

  As described above, the hash data calculation process shown in FIG. 17 shows an operation example realized by the hash data calculation unit 1420 or the hash data calculation unit 1223. Further, the hash data comparison processing shown in FIG. 18 shows an operation example realized by the hash data comparison unit 1430 or the hash data comparison unit 1224.

  Next, the operation (operation 1) of the present embodiment will be described in detail with reference to FIGS. 2, 11, 13 to 18, and specific examples (FIGS. 19 to 21). Here, a specific example will be mainly described, and a part similar to the above-described operation will be omitted.

  As a premise, the full backup data shown in FIG. 19A is stored in the full backup data storage unit 1110 of FIG. 2, and the hash data corresponding to the full backup data of FIG. 19A is the working area 1300 of FIG. It is assumed that it is stored in An example of the contents of hash data is shown in FIG.

  When a differential backup is instructed by a user operation or the like, the boot image transmission unit 1600 in FIG. 2 transmits the boot image 1220 to the managed machine and restarts the managed machine (step S12 in FIG. 12). .

The restarted managed machine is started with the boot image 1220 of FIG. 11 (step S22 of FIG. 12), and the differential backup process (FIG. 13) is started (step S23 of FIG. 12).
When the differential backup process is started, the data transmitting / receiving unit 1221 in FIG. 11 receives the hash data in FIG. 19B from the management server 1000 (step A1 in FIG. 13).
After receiving the hash data of FIG. 19B, the difference data creation means 1222 of FIG. 11 inputs “disk area of the managed machine” as the calculation target area and “512K” (“K” is kilobytes) as the size. The hash data calculation means 1223 is called as (step A2 in FIG. 13). An example of hash data calculated by the hash data calculation unit 1223 is shown in FIG.

  After calculating the hash data of the disk area of the managed machine (hash data in FIG. 20A), the difference data creation unit 1222 uses the hash data comparison unit 1224 to compare the hash data in FIG. The hash data of (a) are compared (step A3 in FIG. 13). Since the hash values of the blocks starting with addresses 0 and 512K are different, the difference data creation unit 1222 creates these as difference data (difference data in FIG. The difference data is transmitted to (step A4 in FIG. 13). FIG. 20B shows an example of difference data created. In this example, difference data is created for 512K starting from address 0 and 512K starting from address 512.

When the data transmission / reception unit 1500 receives the difference data from the managed machine, the management server 1000 in FIG. 2 stores the difference data in the difference data storage unit 1120 and starts the division / combination processing (FIG. 14).
The division efficiency calculation unit 1410 in FIG. 2 includes the difference in FIG. 20B stored in the differential data storage unit 1120 out of the full backup data in FIG. 19A stored in the full backup data storage unit 1110. An area corresponding to the data (previous area) is extracted (step C1 in FIG. 15).

  After dividing the previous area, the division efficiency calculation unit 1410 causes the hash data calculation unit 1420 to calculate hash data for the previous area using a small block size. Specifically, the hash data calculation means 1420 calls the hash data calculation means 1420 with “previous area” as the calculation target area and “256K” as the size as input (step C2 in FIG. 15), and the hash data (step C2). Create the previous area). FIG. 21A shows an example of created hash data (previous area). After the hash data calculation unit 1420 creates the hash data, the division efficiency calculation unit 1410 sets “data in the difference data in FIG. 20B” as the calculation target area and “256K” as the size (1/2 times 512K). As an input, the hash data calculation means 1420 is called (step C3 in FIG. 15) to create hash data (data in the difference data). FIG. 21B shows an example of hash data (data in difference data).

  After creating the hash data (previous area) in FIG. 21 (a) and the hash data (data in the difference data) in FIG. 21 (b), the division efficiency calculation means 1410 in FIG. The hash data comparison unit 1430 is called by inputting the hash data (previous area) of FIG. 21 and the hash data (data in the difference data) of FIG. 21B (step C4 of FIG. 15). The hash data comparison unit 1430 detects that the hash values are different only in blocks starting with addresses 256K and 768K. Therefore, since the difference data amount is 512K (step C5 in FIG. 15), it is determined that “the difference data size is reduced” (step C8 in FIG. 15).

As a result of the division efficiency calculation process, the amount of difference data decreases. Therefore, it is determined that the division size is 256K, and the difference data in FIG. 20 is converted to the difference data shown in FIG. (Step B9 in FIG. 14) to create “full backup data” as the calculation target area and “256K” as the size as input to create hash data (hash data to be transmitted to the next managed machine) ( Step B10 in FIG. 14 is stored in the working area 1300, and the process is terminated.
As described above, the divided block size at the next differential backup is executed at 256K.

  Next, the operation (operation 2) of the present embodiment will be described in detail with reference to FIGS. 2, 11 to 18 and specific examples (FIGS. 19 to 20 and FIG. 22).

As a premise, the full backup data of FIG. 19A is stored in the full backup data storage unit 1110 of FIG. 2, and the hash data corresponding to the full backup data of FIG. 19A is stored in the working area 1300 of FIG. Assume that it is stored. An example of the contents of hash data is shown in FIG.
Further, to simplify the explanation, it is assumed that the difference data is determined not to decrease in step B2 of FIG.

Since the difference data amount does not decrease as a result of the division efficiency calculation process, the coupling efficiency calculation unit 1440 in FIG. 2 converts the full backup data in FIG. 19A stored in the full backup data storage unit 1110 to 1024K (512K). Is divided into blocks having a size of 2), and a block including the data in the differential data (a portion of 0 to 1024K of the full backup data in FIG. 22A) is extracted (step D1 in FIG. 16).
The portion of 0 to 1024K of the full backup data in FIG. 22A and the difference data amount in FIG. 22B are compared (step D2 in FIG. 16), and “the data size does not increase” because the sizes are the same. (Step D5 in FIG. 16). Specifically, in the case of FIG. 22A, one block is 1024K, but the area extracted as difference data (FIG. 22B) corresponds to the area of one block. For this reason, the amount of difference data does not increase.

  Since the difference data amount does not increase as a result of the coupling efficiency calculation process, it is determined that the division size is 1024K, and the difference data of FIG. 20 is converted into the difference data of FIG. 20 in FIG. (Step B5 in FIG. 14), then the hash data (the hash data to be sent to the managed machine next time) is created by inputting “full backup data” as the calculation target area and “1024K” as the size ( Step B6 in FIG. 14 is stored in the working area 1300, and the process is terminated.

  As described above, the divided block size at the next differential backup is executed at 1024K.

(Other embodiments)
In the first embodiment, the division / combination unit 1400 has been described with respect to the case where the block size is divided into half or combined twice, but the block size may be changed using other numerical values. For example, the dividing / combining unit 1400 calculates the difference data amount between the case where the decrease value smaller than the current value is used for the block size and the case where the multiple value of the current value is used for the block size. The block size may be calculated according to The decrease value is preferably a divisor of the current value. The multiple value is M times the current value (an integer greater than M> 0).

  Specifically, the setting method is the same as the logic shown in FIG. 8, and the dividing / combining means 1400 sets the block size to the decrease value when the difference data amount decreases when the decrease value is used. To do. When the difference data amount is the same when using the decrease value and when using the multiple value, the block size is set to the multiple value. If the difference data amount is the same when the decrease value is used and the difference data amount increases when the multiple value is used, the block size is not changed.

  In the first embodiment, the case where the full backup data is a full copy of data from the address 0 of the disk area of the managed machine to the maximum address N of the disk has been described. The present invention can be applied even when copying. In such a case, the full backup data means complete backup data of a disk area determined to be backed up in advance.

  Moreover, when each means demonstrated in Embodiment 1 is implement | achieved using a program, a program can be recorded on a computer-readable recording medium. In addition, each storage area described in the first embodiment can be secured in the storage area of the management server 1000 or the management target machine by a program.

  As an example, the program causes the computer to execute the following process. Processing for backing up the full backup data of the managed machine to the full backup data storage unit 1110. A process of storing hash data including a block size for dividing the full backup data into a plurality of blocks and a plurality of hash values corresponding to each block in a working area (hash data storage means) 1300. A process of sending a boot image that causes the managed machine to extract the difference data between the full backup data and the current data using hash data when the managed machine starts backup, and restarting the managed machine. Processing to notify the managed machine of hash data and extract difference data. Processing to calculate the block size used for the next backup using full backup data and differential data. Processing to store the latest full backup data reflecting the differential data in the full backup data storage means. A process of calculating a hash value corresponding to each block obtained by dividing the latest full backup data using the calculated block size. Processing for storing the calculated block size and the calculated hash value in the hash data storage means.

As described above, according to any one of the above embodiments, the following effects can be obtained.
The first effect is that the cost of system construction can be reduced because there is no need to have a difference calculation mechanism in the managed machine. The reason is that the management server has a differential calculation mechanism (backup module) and transmits it to the managed machine when executing differential backup.

  The second effect is that the difference data is calculated on the management server after the full backup is collected, or the difference is calculated on the managed machine by sending the previously collected full backup data to the managed machine. Compared to the above, the network load is reduced. The reason is that only the hash data calculated in advance is transmitted to the managed machine, and the difference calculation is performed based on it.

  A third effect is that the difference data amount is optimized (minimized) as the backup is collected for the managed machine that exhibits a specific access tendency. The reason is that, after execution of differential backup, an appropriate value for the block size for differential calculation is calculated, and if it is not an appropriate value, the block size is automatically changed.

  According to the present invention, it is possible to perform differential backup reasonably without providing a differential calculation mechanism on a managed machine. It can also be applied to backup of managed machines that are characterized by disk access (not random).

  In addition, this invention is not limited to embodiment shown above. Within the scope of the present invention, it is possible to change, add, or convert each element of the above-described embodiment to a content that can be easily considered by those skilled in the art.

1000 Management server 1100 Backup data storage area 1110 Full backup data storage means 1120 Differential data storage means 1200 Boot image storage area 1220 Boot image 1221 Data transmission / reception means 1222 Differential data creation means 1223 and 1420 Hash data calculation means 1224 and 1430 Hash data comparison means 1300 Working Area 1400 Division / Combination Unit 1410 Division Efficiency Calculation Unit 1440 Connection Efficiency Calculation Unit 1500 Data Transmission / Reception Unit 1600 Boot Image Transmission Unit 2000 Management Network 3000-5000 Managed Machine

Claims (10)

A management server that backs up data on managed machines,
Full backup data storage means for storing full backup data of the managed machine;
Hash data storage means for storing hash data including a block size for dividing the full backup data into a plurality of blocks and a plurality of hash values corresponding to each block;
Boot image storage means for storing a boot image that causes the managed machine to extract difference data between the full backup data and current data using the hash data;
Transmitting means for transmitting the boot image to the managed machine when the managed machine starts backup;
A data transmission / reception means for transmitting the hash data to the managed machine after the managed machine is started using the boot image, and receiving differential data extracted by the managed machine;
The block size used for the next backup is calculated using the full backup data and the differential data, the latest full backup data reflecting the differential data is stored in the full backup data storage means, and the calculated block A hash value corresponding to each block obtained by dividing the latest full backup data using a size, and a dividing / combining means for storing the calculated block size and the calculated hash value in the hash data storage means, Management server provided.   The dividing / combining means calculates a difference data amount between a case where a decrease value smaller than a current value is used for the block size and a case where a multiple value of the current value is used for the block size, and the difference data The management server according to claim 1, wherein the block size is calculated according to an amount.   The dividing / combining means sets the block size to the decrease value when the difference data amount decreases when using the decrease value, and uses the multiple value when the decrease value is used. When the difference data amount is the same when the block size is set to the multiple value, the difference data amount is the same when the decrease value is used, and the multiple value is used The management server according to claim 2, wherein the block size is not changed when a difference data amount increases.   4. The dividing / combining means uses a half of the current value as the decrease value, and uses a value obtained by doubling the current value as the multiple value. Management server. A backup method that backs up data on a managed machine using the management server.
The management server
Full backup data storage means for storing full backup data of the managed machine;
Hash data storage means for storing hash data including a block size for dividing the full backup data into a plurality of blocks and a plurality of hash values corresponding to each block;
Boot image storage means for storing a boot image that causes the managed machine to extract difference data between the full backup data and current data using the hash data;
Transmitting means for transmitting the boot image to the managed machine when the managed machine starts backup;
A data transmission / reception means for transmitting the hash data to the managed machine after the managed machine is started using the boot image, and receiving differential data extracted by the managed machine;
The block size used for the next backup is calculated using the full backup data and the differential data, the latest full backup data reflecting the differential data is stored in the full backup data storage means, and the calculated block A hash value corresponding to each block obtained by dividing the latest full backup data using a size, and a dividing / combining means for storing the calculated block size and the calculated hash value in the hash data storage means, Prepared,
The managed machine is
Data transmitting / receiving means for receiving the boot image and the hash data and transmitting the difference data;
A backup method comprising: difference data creating means for calculating hash data of backup target data and creating difference data by comparing with the received hash data. A backup method for backing up data on a managed machine,
Back up the full backup data of the managed machine to the full backup data storage means,
Hash data including a block size for dividing the full backup data into a plurality of blocks and a plurality of hash values corresponding to each block is stored in a hash data storage unit,
Store the boot image storage means in the boot image storage means to extract the difference data between the full backup data and the current data using the hash data to the managed machine,
When the managed machine starts backup, restart the managed machine using the boot image,
Notifying the managed machine of the hash data and extracting the difference data;
Using the full backup data and the differential data, the block size used for the next backup is calculated,
The latest full backup data reflecting the difference data is stored in the full backup data storage means,
Calculate a hash value corresponding to each block obtained by dividing the latest full backup data using the calculated block size,
A backup method for storing the calculated block size and the calculated hash value in the hash data storage means. The block size calculation is
Calculating a difference data amount between a case where a decrease value smaller than a current value is used for the block size and a case where a multiple value of the current value is used for the block size;
The backup method according to claim 6, wherein the block size is calculated according to the difference data amount. The block size calculation is
When the difference data amount decreases when the decrease value is used, the block size is set to the decrease value,
When the difference data amount is the same when using the decrease value and when using the multiple value, the block size is set to the multiple value,
8. The block size is not changed when the difference data amount is the same when the decrease value is used and the difference data amount increases when the multiple value is used. Backup method. The block size calculation is
9. The backup method according to claim 7, wherein a half of the current value is used as the decrease value, and a value obtained by doubling the current value is used as the multiple value. A program that implements backup that backs up data on managed machines,
On the computer,
A process of backing up the full backup data of the managed machine to a full backup data storage means;
Processing for storing hash data including a block size for dividing the full backup data into a plurality of blocks and a plurality of hash values corresponding to each block in a hash data storage unit;
When the managed machine starts backup, it transmits a boot image that causes the managed machine to extract difference data between the full backup data and current data using the hash data, and Process to restart,
Processing for notifying the managed machine of the hash data and extracting the difference data;
Using the full backup data and the difference data, a process of calculating a block size used for the next backup;
A process of storing the latest full backup data reflecting the difference data in the full backup data storage means;
A process of calculating a hash value corresponding to each block obtained by dividing the latest full backup data using the calculated block size;
A program for executing a process of storing the calculated block size and the calculated hash value in the hash data storage means.

Images