JP5702652B2 - Memory synchronization method, active virtual machine, standby virtual machine, and memory synchronization program - Google Patents

Description

  The present invention relates to a memory synchronization method, an active virtual machine, a standby virtual machine, and a memory synchronization program, and in particular, in a virtual machine synchronization mechanism having a redundant configuration using two or more physical machines, When the standby storages match (for example, using shared storage), a memory synchronization method for synchronizing the memory states of the active and standby virtual machines, the active virtual machine, and the standby virtual machine And a memory synchronization program.

  In order to realize a highly reliable system by preparing two or more physical machines and making them redundant, it is necessary to synchronize the systems.

  As a method for this, there is a system that transfers the execution state of a virtual machine to a standby physical machine while minimizing service stop time of the virtual machine operating on the active physical machine (for example, non-patent literature) 1). Since this system transfers the execution state of the virtual machine without stopping the virtual machine, the execution state is constantly updated. Therefore, if a change occurs in the already transferred virtual machine memory, it is necessary to transfer it again. When the execution state of the virtual machine to be transferred becomes sufficiently small, the active virtual machine is stopped and the remaining data is transferred to the standby system.

  In addition, there is a system in which the execution state of the primary and secondary virtual machines is always kept the same by synchronizing the execution state of the virtual machine triggered by network output or storage writing (for example, see Non-Patent Document 2). Since this system does not synchronize the storage used by the virtual machine, it uses shared storage.

  The system will be described below.

  FIG. 1 shows the configuration of a conventional synchronization system.

  Both the active system 1A and the standby system 1B have a virtual machine control unit 10, and each virtual machine control unit 10 has a synchronization control unit 11, which occurs in the execution state of a virtual machine operating on an active physical machine. The change is transferred to the physical machine of the standby system 1B and applied to the standby virtual machine, so that the execution states (CPU, memory, device) of the virtual machines 20A and 20B are always kept in the same state. Although the figure shows an example in which the shared storage 2 is shared by the hardware 30A and 30B of the active system 1A and the standby system 1B, the configuration for keeping the storage 2 in the same state is shown in FIG. The configuration is not limited, and for example, the active system and the standby system may use separate storages to perform processing that always keeps the same.

  The active system 1A and the standby system 1B perform the operations shown in FIGS. 2 and 3, respectively.

  FIG. 2 is a flowchart of an active system synchronization control unit according to the prior art.

  When the synchronization control unit 11A of the virtual machine control unit 10A of the active system 1A detects that writing has occurred in the storage 2, it starts synchronization (step 110). The difference memory from the previous synchronization time is acquired, and the difference of the execution state (CPU, difference memory, device) of the active virtual machine 20A is transmitted to the standby synchronization control unit 11B (step 120). It is determined whether or not all data reception completion notification has been transmitted (step 130).

  FIG. 3 is a flowchart of a standby synchronization control unit according to the prior art.

  When the synchronization control unit 11B of the virtual machine control unit 10B of the standby system 1B receives the difference in the execution state of the active machine (step 210), the received difference is reflected in the execution state of the standby system virtual machine 20B (step 220). ) An all data reception completion notification for notifying that all data has been received is transmitted to the operation system 10A on the transmission side (step 230).

  In addition to the above technology, the paper introduced that the execution state and storage of the primary and secondary virtual machines can always be kept the same by synchronizing before issuing the network output ( For example, refer nonpatent literature 3).

  Further, there is a technique of transferring a difference and synchronizing when there is a discrepancy between data in a master and a slave storage device (see, for example, Patent Document 1).

Japanese Patent No. 4472959

Clark, C., Fraser, K., Hand, S., Hansen, JG, E., Limpach, C., Paratt, I. and Warfield, A .: Live Migration of Virtual Machines, 2nd USENIX Symposium on Networked Systems Design and Implementation, Boston, MA, USA, pp. 273-286. Yoshiaki Tamura, Kari Yanagisawa, Koji Sato, Satoshi Moriai: Kemari: Fault Tolerant Clustering by Synchronization Between Virtual Machines, IPSJ Transactions, Vol. 3, No.1, pp. 13-24 (2010). B. Curry., G. Lefebvre., D. Meyer., M. Feeley., N. Hutchinson., A. Warfield .: Remus: High Availability via Asynchronous virtual Machine replication, 5th USENIX Symposium on Networked Systems Design and Implementation, pp. 161-174 (2008).

  However, since the technique of Non-Patent Document 1 described above is a method of transferring memory data itself in order to match the memory of a virtual machine that has changed due to storage read, there is a problem that the transfer amount is large. .

  Further, the technique of Non-Patent Document 2 has a problem that the transfer amount is large because the memory data itself is transferred in order to match the memory of the virtual machine in which the change has occurred by reading the storage.

  Further, the technique of Non-Patent Document 3 has a problem that the transfer amount is large because the memory data itself is transferred in order to match the memory of the virtual machine in which the change has occurred by reading the storage.

  Further, the technique of Patent Document 1 has a problem that the transfer amount is large because the master transfers the data itself in the storage device to the slave.

  The present invention has been made in view of the above points, and is a memory synchronization method capable of reducing the amount of data transfer when the active and standby storages match (for example, using shared storage). It is another object of the present invention to provide an active virtual machine, a standby virtual machine, and a memory synchronization program.

In order to solve the above problem, the present invention has a shared storage in the active system and the standby system, and operates on the physical contents of the virtual machine that operates on the active physical machine and on the standby physical machine. A memory synchronization system that synchronizes the contents of a virtual machine's memory at an arbitrary timing,
In the operational virtual machine,
When a read command is generated in the shared storage, the read command capturing unit stores a read position, a read position, and a read size of the read command in a storage unit, and
The difference information transfer means, when a write command occurs in the shared storage, a difference information transfer step of acquiring the difference information of the execution state from the previous synchronization time and transmitting it to the standby virtual machine;
Synchronous data transfer means acquires the read position, the read position, and the read size from the storage means, and transfers the synchronous data transfer step to the standby system;
In the standby virtual machine,
A difference information reflecting step in which the difference information reflecting means reflects the difference information of the active virtual machine received from the active virtual machine to the standby virtual machine;
A synchronization step in which synchronization means reads the data of the shared storage specified by the read position into the standby virtual machine memory specified by the read position received from the active virtual machine by the read size; ,I do.

  As described above, according to the present invention, when the shared storage is used in the active system and the standby system, the difference in the execution state of the active virtual machine and the contents of the storage read command (read position, read position, read size) Since the virtual machine memories of the active system and the standby system are made to coincide with each other by transferring to the standby system, it is not necessary to transfer the data itself, so that the transfer amount can be reduced.

  In the standby system, since the shared storage is read based on the read command acquired from the active system, the states of the memories can be matched.

It is a block diagram of the conventional synchronous system. It is a flowchart of the active system synchronous control part by a prior art. It is a flowchart of the standby system synchronous control part by a prior art. It is a whole flowchart of the active system synchronous control part in one embodiment of this invention. It is a detailed flowchart of S320 in one embodiment of this invention. It is a detailed flowchart of S340 in one embodiment of the present invention. It is a flowchart of the synchronous control part of the standby system in one embodiment of the present invention. It is a figure (the 1) which shows the structure of the physical machine in one embodiment of this invention. It is FIG. (2) which shows the structure of the physical machine in one embodiment of this invention. It is a figure which shows the experimental environment for evaluation. This is the data transfer amount per synchronization between the present invention and the prior art.

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

In the present invention, synchronization and is an the contents of the memory of the virtual machines running on the operational system of the physical machine, causes the contents of the memory of the virtual machines running on the standby physical machines matched at an arbitrary timing process , Shall be conducted continuously.

  The system configuration of the present invention is the same as that of FIG. 1 described above, but the processing performed by the synchronization control unit 11 is different. In the following, since the application is operating in the active virtual machine 20A, the execution state is constantly updated, and the application is not operating in the standby virtual machine 20B. Is assumed to be unchanged. In the following embodiment, both the active system 1A and the standby system 1B use the shared storage 2.

  First, processing of the synchronization control unit 11A of the active system 1A will be described.

  FIG. 4 is an overall flowchart of the active synchronization control unit according to the embodiment of the present invention.

  Step 310) The synchronization control unit 11A determines whether or not a read to the shared storage 2 has occurred, and proceeds to Step 320 if it has occurred.

  Step 320) The synchronization control unit 11A captures the reading from the shared storage 2, and stores the read command in a storage unit (not shown) such as a main storage device. Details of this processing will be described later with reference to FIG.

  Step 330) Next, the synchronization control unit 11A determines whether an active storage write has occurred, and starts synchronization if it detects the occurrence. Note that the processing for this is the same as the method of performing the synchronization processing triggered by writing to the shared storage 2 as in FIG. 2 (Non-Patent Document 2).

  Step 340) The synchronization control unit 11A transfers data necessary for synchronizing the active system 1A and the standby system 1B to the standby system 1B. Details of the processing will be described later with reference to FIG.

  Next, the process of step 320 will be described.

  FIG. 5 is a detailed flowchart of step 320 in one embodiment of the present invention.

  Step 410) The synchronization control unit 11A captures reading from the storage 2 by the virtual machine 20A.

  Step 420) Upon detecting the storage read, the synchronization control unit 11A stores the storage read command in a storage unit (not shown) such as a main storage device. Here, the storage read command information to be stored includes a read position (for example, a virtual machine page address), a read position (for example, a sector number of the storage 2), a read size, and the like.

  Next, the process of step 340 in FIG. 4 will be described. Information transmitted to the standby synchronization control unit 11B is different from that of Non-Patent Document 2.

  FIG. 6 is a detailed flowchart of step 340 in one embodiment of the present invention.

  Step 510) The synchronization control unit 11A acquires the difference memory from the previous synchronization from the active virtual machine 20A. However, memory that has changed due to storage read is excluded. The difference (CPU, difference memory, device) of the execution state of the active virtual machine 20A is transmitted to the synchronization control unit 11B of the standby system 1B.

  Step 520) The synchronization control unit 11A takes out the storage read command (read data position, read position, read size) stored in the storage means (not shown) in Step 320, and sends it to the synchronization control unit 11B of the standby system 1B. Send. In addition, this data may consist of a plurality.

  Step 530) It is determined whether a reception completion notification for all data has been transmitted from the standby system 1B on the receiving side, and if it has been transmitted, the process proceeds to Step 540.

  Step 540) The storage read command stored in the storage means (not shown) in Step 520 is discarded. This is a process for preventing the transfer of the same data at the next synchronization.

  Next, processing of standby system 1B will be described.

  FIG. 7 is a flowchart of the standby-system synchronization control unit according to the embodiment of the present invention.

  Step 610) The synchronization control unit 11B performs synchronization data such as a difference in execution state of the virtual machine 20A of the active system 1A transmitted from the active system 1A by the processing of the above-described steps 510 and 520 and a read command for the storage 2 and the like. If it is received, the process proceeds to step 620.

  Step 620) The received difference in the execution state of the virtual machine 20A of the active system 1A is reflected in the execution state of the virtual machine 20B of the standby system 1B.

  Step 630) When the standby control storage 2 data and the storage data read command (read position, read data position, read size) are input, the synchronization control unit 11B on the main storage device designated by the read position Data of the storage 2 designated by the position of the read data is read into the virtual machine memory for the designated size. Thereby, the virtual machine memory of the active system 1A can be matched with the virtual machine memory of the standby system.

  Step 640) A reception completion notification of all data from the active system side is transmitted to the active system 1A.

  As described above, in the present invention, since the storage read command is stored in the storage unit of the active system 1A in Step 420, it can be used in Step 520 and Step 630. That is, even if a change occurs in the memory of the virtual machine of the active system 1 by reading the storage data, the same storage data exists in the storage 2 and therefore the data itself does not need to be transferred. Since the data size of the storage read instruction transferred instead of the virtual machine memory is sufficiently smaller than the data of the virtual machine memory itself, the transfer amount can be reduced.

  Further, since the standby system 1B reads from the storage 2 similar to the active system 1A based on the storage data read command in step 630, the states of the memories can be matched.

  In the above embodiment, the case where the active machine and the standby machine are 1: 1 is shown, but an example in which the active machine and the standby machine are N: 1 or 1: N is shown below.

  FIG. 8 is a diagram (part 1) illustrating the configuration of the physical machine according to the embodiment of the invention.

  In the configuration shown in the figure, the active machine A and the standby machine B share the storage a, and the standby machine B shares the storage b.

  Regarding the configuration of the physical machine, it is possible to adopt a configuration in which the number of active machines: the number of standby machines = N: 1. One of the virtual machines of the standby machine shown in the figure is synchronized with the virtual machine of the operational machine A, and the other virtual machine is synchronized with the operational machine B. However, regarding the configuration of virtual machines, the number of active virtual machines: the number of standby virtual machines = N: 1, and one or more standby virtual machines serving as backups for the active virtual machines are required.

  When there are a plurality of standby virtual machines at the backup destination of a certain active virtual machine, it is necessary to select which one will be the new active system after the failure of the active system.

  Further, the configuration for keeping the storage in the same state is not limited to the example of FIG. For example, the active system and the standby system may use different storages and perform processing to keep them always the same.

  FIG. 9 is a diagram (part 2) illustrating the configuration of the physical machine according to the embodiment of the invention.

  In the configuration shown in the figure, the active system and standby systems A and B share storage.

  As shown in the figure, regarding the physical machine configuration, it is possible to adopt a configuration where the number of active machines: the number of standby machines = 1: N. Both virtual machines on the standby machines A and B shown in the figure are synchronized with the active virtual machine. However, regarding the configuration of the virtual machines, one or more standby virtual machines that are backups of the active virtual machines are required, ie, the number of active virtual machines: the number of standby virtual machines = 1: N.

  When there are a plurality of standby virtual machines at the backup destination of a certain active virtual machine, it is necessary to select which one will be the new active system after the failure of the active system.

  Further, the configuration for keeping the storage in the same state is not limited to the examples of FIGS. 8 and 9. For example, the active system and the standby system may use different storages and perform processing to keep them always the same. Good.

  Next, the results of performance evaluation of the present invention and the prior art will be shown.

  FIG. 10 shows an experimental environment for evaluation.

  In the figure, for the sake of evaluation, two machines serving as experimental environments are prepared, one being an operation system and the other being a standby system. File copy is executed on the active guest OS and a load is applied.

  FIG. 11 shows the data transfer amount per synchronization between the present invention and the prior art.

  The data in the figure is an average value of the data transfer amount generated when the active system and the standby system are synchronized 100 times. When the present invention is used, the following effects can be obtained as compared with the prior art (Non-Patent Document 2). Since the synchronization timing of the present invention and Non-Patent Document 2 are the same, there is no change in the number of synchronizations that occur.

  -Overall (other + memory information changed by storage read from 571 KB to 494 KB, reducing the data transfer amount by 13%. Note that the above "other" means the virtual machine execution state (CPU, Device, memory) minus the memory data size that has changed due to storage read.

  -The memory information changed due to the storage read was changed from 77 KB to 0.14 KB, and the data transfer amount was reduced by 99%. The above-mentioned “memory information changed by storage read” is memory data in the case of the prior art (Non-Patent Document 2), and is a storage read instruction in the case of the present invention.

  The operations of the active and standby synchronization control units can be constructed as a program, and can be installed and executed on a physical machine used as a virtual machine.

  The present invention is not limited to the above-described embodiments, and various modifications and applications are possible within the scope of the claims.

1A Operation system (data transmission side)
1B Standby system (data receiving side)
2 Shared storage 10A Virtual machine controller (active)
10B Virtual machine control unit (standby system)
11A Synchronization control unit (operational system)
11B Synchronization control unit (standby system)
20A virtual machine (operational)
20B Virtual machine (standby system)
30A hardware (operational)
30B hardware (standby system)

Claims (5)

The active and standby systems have shared storage, and the contents of the virtual machine memory running on the active physical machine and the virtual machine memory contents running on the standby physical machine are synchronized at any time A memory synchronization system,
The active virtual machine is
When a read command is generated in the shared storage, a read command capturing unit that stores a read position, a read position, and a read size of the read command in a storage unit;
When a write command occurs in the shared storage, difference information transfer means for acquiring difference information of an execution state from the previous synchronization time and transmitting the difference information to the standby virtual machine;
Synchronous data transfer means for acquiring the read position, the read position, and the read size from the storage means and transferring them to the standby system;
Have
The standby virtual machine is
Differential information reflecting means for reflecting the difference information of the active virtual machine received from the active virtual machine to the standby virtual machine;
Synchronization means for reading the data of the shared storage specified at the read position into the standby virtual machine memory specified at the read position received from the active virtual machine for the read size;
A memory synchronization system comprising: A shared storage in a production system and the standby system, synchronized at any time the contents of the memory of the virtual machines running on the content of the memory of the virtual machine standby physical machines that operate on a production system of the physical machine A memory synchronization method,
In the operational virtual machine,
When a read command is generated in the shared storage, the read command capturing unit stores a read position, a read position, and a read size of the read command in a storage unit, and
The difference information transfer means, when a write command occurs in the shared storage, a difference information transfer step of acquiring the difference information of the execution state from the previous synchronization time and transmitting it to the standby virtual machine;
Synchronous data transfer means acquires the read position, the read position, and the read size from the storage means, and transfers the synchronous data transfer step to the standby system;
In the standby virtual machine,
A difference information reflecting step in which the difference information reflecting means reflects the difference information of the active virtual machine received from the active virtual machine to the standby virtual machine;
A synchronization step in which synchronization means reads the data of the shared storage specified by the read position into the standby virtual machine memory specified by the read position received from the active virtual machine by the read size; ,
And a memory synchronization method. A shared storage in a production system and the standby system, synchronized at any time the contents of the memory of the virtual machines running on the content of the memory of the virtual machine standby physical machines that operate on a production system of the physical machine An operational virtual machine in a synchronous system,
When a read command is generated in the shared storage, a read command capturing unit that stores a read position, a read position, and a read size of the read command in a storage unit;
When a write command occurs in the shared storage, difference information transfer means for acquiring difference information of an execution state from the previous synchronization time and transmitting the difference information to the standby virtual machine;
Synchronous data transfer means for acquiring the read position, the read position, and the read size from the storage means and transferring them to the standby system;
An operational virtual machine characterized by comprising: A shared storage in a production system and the standby system, synchronized at any time the contents of the memory of the virtual machines running on the content of the memory of the virtual machine standby physical machines that operate on a production system of the physical machine A standby virtual machine in a synchronous system,
At the time when the write command of the virtual machine該運for system received from the virtual machine of the active system occurs, the difference information reflecting means for reflecting the difference information corresponding execution state from the previous synchronization to the virtual machine of the standby system ,
The read position, read position, and read size when a read command is generated from the active virtual machine is received, and the shared virtual machine memory specified at the read position is stored in the standby virtual machine memory specified at the read position. Synchronization means for reading storage data for the read size;
A standby virtual machine characterized by comprising:   The memory synchronization program for functioning a computer as a virtual machine of Claim 3 or 4.

Images