VIRTUAL MACHINE-BASED ON-DEMAND PARALLEL DISASTER RECOVERY SYSTEM AND THE METHOD THEREOF

• US 20100115332A1
• Filed: 06/24/2009
• Published: 05/06/2010
• Est. Priority Date: 11/06/2008
• Status: Active Grant

• Current Assignee: Hongliang Yu, Weimin Zheng
• Original Assignee: Hongliang Yu, Weimin Zheng
• Inventors: Yu, Hongliang, Zheng, Weimin
• Granted Patent: US 8,161,321 B2
• Time in Patent Office: Days
• US Class Current: 714/6
• CPC Class Codes: G06F 11/1451 by selection of backup contents
• CPC Class Codes: G06F 11/1461 Backup scheduling policy
• CPC Class Codes: G06F 11/1464 for networked environments
• CPC Class Codes: G06F 11/1466 to make the backup process non-disruptive
• CPC Class Codes: G06F 11/1469 Backup restoration techniques
• CPC Class Codes: H04L 67/1097 for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS]

Abstract

Embodiments disclosed herein related to a virtual machine-based on-demand parallel disaster recovery system and a method thereof. By integrating context of processes of virtual machine, the system and method build a process tree related to applications and support the operation of freezing the context and intercept memory image consistent with the virtual machine environment, based on which, a method for on-demand parallel recovery is provided. In the method, while monitoring file system data that processes of virtual machine are to visit by a kernel intercept unit, recovery of related data blocks of the applications is conducted in the background in order to make applications and recovery procedure of the virtual machine run simultaneously. The recovery procedure is divided into two phases, recovering memory image first and then recovering file system or volume data. Through the procedure, applications can be recovered precisely to the running state when they are backup. The method realizes the scheduling of multiple backup tasks and recovery tasks in client and realizes the prefetching of the recovered data. The embodiments described herein consume less time for recovery, enhances the reliability of the protected applications, has a low cost and practical value; and it is transparent to users.

Claims

1. A virtual machine-based on-demand parallel disaster recovery system comprises a disaster recovery center and a client connected to each other;

the disaster recovery center is made of multiple servers to compose a storage sharing cluster, the disaster recovery center comprises;

a disaster recovery service guard unit which used to monitor ports and to receive backup or recover requests from the client and correspondingly to backup or recover the data;

a backup data management unit used to generate and store corresponding memory image volumes and data image volumes in response to backup requests from the client;

a cluster volume management unit which virtualizes a sharing storage pool for all the nodes to use and stores and manages memory image volumes and data image volumes in a storage pool in form of logic volume; and

an index file management unit used to search and locate the memory image volumes and data image volumes in the storage pool;

the client comprises;

a virtual machine environment building unit used to build an independent process group and sessions in a host system of the client and to encapsulate applications and corresponding resources protected by users and then to build the virtual machine environment where processes of the client are located;

a kernel interception unit used to monitor each read-write request in the virtual machine environment and to search and maintain a recovery data bitmap that records the information related to the recovered data;

a virtual machine freezing unit used to search the recovery data bitmap when the kernel interception unit has monitored that there is read-write request of processes of the virtual machine environment;

if the data to be read and written is not recovered, the operation of freezing the virtual machine environment is initiated in core state;

a virtual machine unfreezing unit used to initiate the operation of unfreezing the virtual machine after the data to be read and written is written into the local storage media in the client in order to make each process in the virtual machine environment resume operation; and

a request forward guard unit used to communicate with the disaster recovery center, to initiate backup or recover request, sending and receiving data and to centrally manage data request and schedule to realize the disaster recovery.

2. The virtual machine-based on-demand parallel disaster recovery system according to claim 1, wherein the request forward guard unit in the client of the system comprises a backup subunit and a recovery subunit, wherein:

the backup subunit is used to receive backup requests sent from users, to generate backup tasks, to establish a connection with the disaster recovery center according to the need of backup tasks, to forward the subsequent requests of the backup task to the servers and to centrally schedule the backup requests and control flow of them according to the priority of each backup task; and

the recovery subunit is used to receive the recover requests sent from users and the kernel interception unit, to generate recovery tasks, to establish a connection with the disaster recovery center according to the need of the recovery tasks, to record the data sent from disaster recovery center and write the data into the local storage media, to execute callback function to complete data recovery in response to the recover requests from the kernel interception unit simultaneously, and to centrally schedule the recover requests and control flow of them according to the priority of each recovery task.

3. The virtual machine-based on-demand parallel disaster recovery system according to claim 1, wherein the disaster recovery center in the system further comprises a cluster volume guard unit communicating with the cluster volume management unit, the cluster volume guard unit is used to maintain the state of the sharing storage cluster and to ensure the number of effective nodes in the sharing cluster and the synchronization of the modification made to the memory image volumes and data image volumes in the storage pool.

4. The virtual machine-based on-demand parallel disaster recovery system according to claim 1, wherein the disaster recovery center in the system further comprises a cluster volume drive unit and a physical device drive unit, the cluster volume drive unit is used to maintain the mapping information of logic data blocks and physical data blocks of the memory image volumes and data image volumes and to schedule and distribute the read-write orders of each logic volume, the layer under the cluster volume drive unit is a physical device drive unit connecting to the sharing storage pool through the storage area network.

5. A virtual machine-based on-demand parallel disaster recovery method utilizing the said system according to claim 1, the method comprising:

s1.0, creating the virtual machine environment in the client and containing the protected applications in the virtual machine environment;

s1.1, generating the memory image volumes and data image volumes of the virtual machine environment at the designated time point according to the backup request of the client and completing the backup of data;

s1.2, the client initiates the recovery procedure, it generates recovery tasks according to recover requests that users initiate and sends recover requests to the disaster recovery center, and simultaneously informs the kernel interception unit of initializing the recovery data bitmap;

s1.3, the disaster recovery center searches and locates the memory image volumes of the virtual machine environment at the designated time point according to the recover request and sends the memory image volumes to the client;

s1.4, the client captures the memory image of the virtual machine environment at the designated time point according to the received memory image volumes, then rebuilds the memory image to recover the normal running state of the virtual machine environment at the designated time point, and executes step s1.5 and step s1.5′

simultaneously;

s1.5, opening the intercepting function of read-write request of the kernel generic block device in the client;

searching the recovery data bitmap firstly when processes in the virtual machine environment visit the file system;

if data to be read and written have been recovered, visiting the file system directly, otherwise executing step s1.6;

s1.6, the client initiates the operation of freezing the virtual machine environment;

the kernel interception unit issues the recover requests of the read-write data;

s1.7, hanging the recover requests in step s1.6 in the array of the recovery tasks built in step s1.2 and endowing the recover requests with the highest priority and then waiting for the responses from the disaster recovery center;

s1.8, after receiving the read-write data from the disaster recovery center, the client renews the recovery data bitmap by marking the read-write data as “

recovered” and

then writes the marked data into the local storage media;

s1.9, unfreezing the virtual machine to make the processes in the virtual machine environment resume operation; and

s1.5′

, according to the recovery task established in step s1.2, the client continuously sends data block recover requests to the disaster recovery center for the recovery of data blocks in the background;

during recovering, avoiding to repeat recovering data through searching the recovery data bitmap.

6. The virtual machine-based on-demand parallel disaster recovery method according to claim 5, wherein rebuilding the memory image to recover the normal running state of the virtual machine environment at the designated time point conducted in step s1.4 of the method comprises:

s1.4a, deriving a kernel process according to the process trees in the memory image, modifying the kernel stack of the kernel processes to switch the kernel processes to the default self-recovering processes when the system call withdraws;

s1.4b, the kernel processes derive their sub-processes according to the process tree in the memory image;

the said sub-processes are self-recovering processes;

modifying the kernel stack again to construct the withdrawing of the second system call and then switching the kernel processes to the root processes of user state at the designated time point;

s1.4c, deriving new sub-processes in sequence from root to leaves according to the structure of the process tree in the memory image until all the processes in the memory image are derived; and

then setting the state of each derived process as deep-sleep state; and

s1.4d, starting another independent process located outside the virtual machine environment in the client;

according to the orders of users, the independent process resets the states of all the processes in the virtual machine environment in accordance with the records in the memory image to recover the normal running state of the virtual machine environment at the designated time point.

7. The virtual machine-based on-demand parallel disaster recovery method according to claim 5, wherein step s1.0 of building the virtual machine environment in the method comprises:

s1.0a, extending the structure of the processes in the kernel of operation system in the client to make the structure be added with the domain standing for the context of the processes of virtual machine located by the processes;

s1.0b, starting a process of user state to complete the switch between input device and output device and then entering into the kernel state through a system call;

s1.0c, switching the said processes of user states to the user space and executing /sbin/init in the context of the virtual machine to make the processes of user states become the root processes of the virtual machine environment, deriving related sub-processes and then building related system services; and

s1.0d, building the switching mechanism between the context of processes in the local operation system and that of processes of the virtual machine to build the virtual machine environment.

8. The virtual machine-based on-demand parallel disaster recovery method according to claim 5, wherein after issuing the recover request of read-write data in step s1.6, selecting the prefetched data to prefetch by way of submitting the recover request of each data block close to the physical position of the read-write data, wherein, the physical position is determined by the address offset threshold.

9. The virtual machine-based on-demand parallel disaster recovery method according to claim 8, wherein the said address offset threshold is a dynamically optimal value-tuning parameter, the way of tuning is:

when the prefetched data is what the files, which the processes in the virtual machine environment need to visit, need, the absolute value of the said address offset threshold is increased, otherwise it is reduced.

10. The virtual machine-based on-demand parallel disaster recovery method according to claim 8, wherein the priority of the recover request of the prefetched data is lower than that of the read-write data in step s1.6, but higher than that of data blocks sent by the background in step s1.5′

.

11. The virtual machine-based on-demand parallel disaster recovery method according to claim 5, wherein in step s1.5′

, according to the preset recovery strategy, the background continuously sends the recover requests of data blocks to the disaster recovery center to recover data blocks, the said recovery strategy is;

sending the recover requests of data blocks in sequence of the sensitivity level and sending data blocks with high sensitivity level first, the said sensitivity level is set in data blocks, the said sensitivity level is obtained through counting the time of reading and writing of each data block in time window w with a fixed size.

12. The virtual machine-based on-demand parallel disaster recovery method according to claim 5, wherein step s1.8 further comprises compressing the bitmap by the way of tour code after renewing the bitmap.

13. The virtual machine-based on-demand parallel disaster recovery method according to claim 5, wherein in step s1.8, read-write lock mechanism is selected when renewing bitmap, the old data is not deleted when renewing data by the operation of writing until all the read-visits conducted by the old data end, during the process, the operation of reading that newly occurs is located to new data.

14. The virtual machine-based on-demand parallel disaster recovery method according to claim 5, wherein in step s1.2, the client turns each recover request to a two-way link list through link of points, then builds a recover request hash list in memory and puts each recover request into each hash bucket according to its request number.

15. The virtual machine-based on-demand parallel disaster recovery method according to claim 5, wherein in step s1.8 and s1.5, after receiving the recovered data from the disaster recovery center, the client writes the data into the bitmap first and then informs the subsequent procedures that the data has been received, if the data fails to be written, the received data will not be reserved and not be rewritten until the data is written successfully.