System and method for remote mirror data backup over a network

• US 7,464,239 B2
• Filed: 08/02/2006
• Issued: 12/09/2008
• Est. Priority Date: 04/10/2006
• Status: Active Grant

• Current Assignee: Industrial Technology Research Institute
• Original Assignee: Industrial Technology Research Institute
• Inventors: Wey, Shou-Jen, Hwang, Wen-Shyang, Lee, Meng-Yu, Lin, Sheng-Hsiung
• Primary Examiner(s): LANE, JOHN A
• Application Number:
  US11/461,773
• Publication Number:
  US 20070239952A1
• Time in Patent Office: 860 Days
• Field of Search: 711/141, 711/161, 711/162, 711/170, 711/171, 711/172, 711/173
• US Class Current: 711/162
• CPC Class Codes: G06F 11/2058 using more than 2 mirrored copies
• CPC Class Codes: G06F 11/2069 Management of state, configuration or failover
• CPC Class Codes: G06F 11/2074 Asynchronous techniques

Abstract

Disclosed is a method and a system for remote mirror data backup over a network. The system comprises a local domain site and at least one remote domain site, wherein each site has one domain mirror agent and one storage server system thereon. Each storage server system includes one or more storage servers, and one of the domain mirror agents is predefined as the master mirror agent. With such a remote mirror system, the method performs the operations of virtual mirror device allocation, mirror operation, data consistency check, third mirror allocation, and recovery of local storage device. The present invention expands the conventional logical volume management (LVM) to a network environment, effectively improving the backup/recovery capability of remote mirroring.

Claims

1. A system for remote mirror data backup over a network, comprising:

a local domain site having one and only one domain mirror agent responsible for dynamic storage space allocation of a storage server system with one or more storage servers in said local domain site, and communication with other domain sites;

at least one remote domain site, each said remote domain site having one and only one domain mirror agent responsible for dynamic storage space allocation of a storage server system with one or more storage servers in each said remote domain site, and communication with other domain sites; and

wherein one of said domain mirror agents of said local domain site and said at least one remote domain site is pre-defined as a master mirror agent based on a weight value, and said master mirror agent collects information from other said domain mirror agents and dispatches collected information to each said domain mirror agent.

2. The system according to claim 1, wherein said local domain site further includes a physical primary storage device and a virtual secondary storage device, said primary storage device includes primary data to be mirrored;

and said secondary storage device includes mirrored data transmitted and distributed to said at least one remote domain site.

3. The system according to claim 1, wherein said system extends a logical volume management over said network by using a search tree to locate one of said at least one remote domain site for storing mirrored data with a minimum range value of logical extent ranges as index.

4. The system according to claim 1, wherein said storage server system of each said remote domain site further includes a host responsible for communications with said domain mirror agent of a same remote domain site and said one or more storage servers in said same remote domain site are for storing mirrored data distributed to said same remote domain site.

5. The system according to claim 1, wherein said system further includes an examination mechanism to ensure the consistency between mirrored data delivered by said domain mirror agent of said local domain site and mirrored data obtained by said at least one remote domain site.

6. A method of remote mirror data backup over a network having at least one local domain site and at least one remote domain site, each said domain site having a storage server system, said method comprising the steps of:

arranging a domain mirror agent in each said domain site, listing all said domain mirror agents in a domain agent list, and pre-defining a domain mirror agent as a master mirror agent;

dynamically allocating a virtual mirror device in said local domain site to store mirrored data after said domain mirror agent of said local domain site receives a remote mirroring request;

allocating a space mapping table to locate said at least one remote domain site for storing the mirrored data, and conducting a mirroring operation to duplicate the mirrored data from said virtual mirror device to said at least one remote domain site; and

examining the consistency between the mirrored data received by said remote domain sites and the mirrored data transmitted by said domain mirror agent of said local domain site.

7. The method according to claim 6, further including a step of allocating a third mirror to increase the availability and security of the mirrored data.

8. The method according to claim 7, wherein said allocating said third mirror includes a step of selecting a third mirror candidate, said selecting said third mirror candidate includes:

obtaining a weight value φ

_RT according to a function involving a weight value W_RT between said third mirror candidate and said remote domain sites, a weight value W_TL between said local domain site and said third mirror candidate, the numbers of logical extent (LE) ranges N_R that said local domain site assigns to each said remote domain site, the numbers of LE ranges N_T originally responsible by said third mirror candidate; and

selecting said third mirror candidate having a smallest said weight value φ

_RT with each said third mirror candidate being selected only once.

9. The method according to claim 7, wherein said allocating said third mirror includes the steps of:

(a1) selecting a domain mirror agent from a third mirror candidate domain agent list, which is determined by finding those domain mirror agents that are from both said local domain site and a second mirror site;

(b1) checking to see if the domain mirror agent has enough free logical extent (LE) ranges for allocation, if yes, determining the criteria for selecting said third mirror and continuing to step (d1), otherwise, continuing to step (c1);

(c1) searching said domain agent list for another third domain mirror agent, if yes, continuing to step (b1), otherwise terminating said allocating said third mirror;

(d1) establishing a matrix table between said domain mirror agent of said second mirror site and a third mirror agent of said third mirror, and selecting a node having a largest number of LE ranges of said second mirror site as said third mirror candidate wherein a node is a remote mirror host; and

(e1) checking all nodes of said second mirror site to see if they are all allocated as said third mirror, if not, continuing to step (d1), otherwise notifying all nodes to register their LE range allocations and terminating the allocation of said third mirror.

10. The method according to claim 6, further including a step of recovering a storage device for said local domain site.

11. The method according to claim 10, wherein the recovery time of a storage device of said local domain site is determined by the number of logical extent (LE) ranges required by a host of said local domain site, the amount of storage of each LE, and the transmission speeds of connections to all remote domain sites.

12. The method according to claim 10, wherein the recovery of a storage device of said local domain site comprises the steps of:

(a11) specifying said domain mirror agent of said local domain site or another said domain mirror agent to search for a mapping table of said storage device, if found, continuing to step (b11), otherwise continuing to step (c11);

(b11) checking logical extent (LE) ranges of all storage servers listed in said mapping table to see if they are all readable, if yes, continuing to step (d11), otherwise continuing to step (e11);

(c11) specifying another said domain mirror agent from said mirror agent list to search for said mapping table of said storage device, if found, continuing to step (b11), otherwise terminating the recovery of said storage device;

(d11) notifying all nodes to replicate data of their responsible LE ranges to said storage device where a node is a remote mirror host;

(e11) for non-readable LE ranges, searching other domain mirror agents for nodes that have said non-readable LE ranges where a node is a remote mirror host, if not found, terminating the recovery of said storage device, otherwise, continuing to step (f11); and

(f11) checking all LE ranges of all storage servers of said domain mirror agent to see if they are all readable, if yes, continuing to step (d11), otherwise, continuing to step (e11).

13. The method according to claim 6, wherein said master mirror agent collects information regarding the status and transmission speed of network links between each said domain mirror agent and dispatches the collected information back to each said domain mirror agent.

14. The method according to claim 6, wherein said method utilizes a hierarchical domain checking mechanism for the examination of consistency.

15. The method according to claim 14, wherein said hierarchical domain checking mechanism includes the steps of:

XORing all data blocks within a logical extent (LE) range to obtain a LE parity code for each said domain site; and

XORing all LE parity codes to a remote domain global parity code for each said remote domain site.

16. The method according to claim 6, wherein said method utilizes a data variation checking mechanism for the examination of consistency.

17. The method according to claim 16, wherein said data variation checking mechanism includes the steps of:

initializing remote domain time parity codes of all said domain sites;

wherein when some data is varied, a data block having the varied data of a domain site is outputted and XORed with the remote domain time parity code of said domain site; and

wherein when examining data consistency, the remote domain time parity codes of said local domain site and said remote domain site are compared.

18. The method according to claim 6, wherein said dynamically allocating said virtual mirror device in said local domain site includes the steps of:

(a) searching for a corresponding mirror mapping table of a physical primary storage device of said local domain site so as to allocate a virtual secondary storage device, if found, continuing to step (b), otherwise continuing to step (c);

(b) obtaining the status information of storage servers listed in said mirror mapping table and continuing to step (d);

(c) allocating a new mirror mapping table for said domain mirror agent of said local domain site, and checking whether there are enough free logical extent (LE) ranges in said local domain site for allocation to said virtual secondary storage device, if yes, continuing to step (e), otherwise continuing to step (f);

(d) checking if a storage server listed in said mirror mapping table functions properly, if yes, successfully allocating said virtual mirror device and terminating the process of said dynamically allocating said virtual mirror device, otherwise continuing to step (g);

(e) conducting a LE range allocation procedure by said domain mirror agent of said local domain site, notifying said storage server system about the allocated LE ranges to update its allocation status, recording the relevant information in said new mirror mapping table, and continuing to step (b);

(f) searching said domain agent list for another said domain mirror agent, and continuing to step (c); and

(g) discarding said mirror mapping table and the LE range allocation, and continuing to step (c).

19. The method according to claim 6, wherein said mirroring operation includes the steps of:

issuing a first I/O request to a primary storage device of said local domain site;

duplicating said first I/O request to a secondary storage device by a mirror operator;

determining a host ID and a volume group ID (UUID) corresponding to said secondary storage device through a search tree;

converting a logical extent (LE) range of said first I/O request into a corresponding LE range for said secondary storage device;

issuing a second I/O request to said secondary storage device; and

notifying said mirror operator about if a requested I/O operation via said first I/O request and said second I/O request is completed successfully.

20. The method according to claim 6, wherein said allocating a space mapping table includes a step of establishing a search tree, where logical extent (LE) range information is maintained and used as an index to a mirror mapping table for locating a remote domain site where the mirrored data is stored.