METHOD FOR UTILIZING HETEROGENEOUS STORAGE SYSTEMS BY COOPERATING WITH SERVER SIDE STORAGE SOFTWARE

• US 20120117154A1
• Filed: 11/04/2010
• Published: 05/10/2012
• Est. Priority Date: 11/04/2010
• Status: Active Grant

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: KODAMA, Shoji
• Granted Patent: US 8,635,280 B2
• Time in Patent Office: Days
• US Class Current: 709/205
• CPC Class Codes: G06F 16/1844 Management specifically adapted to replicated file systems

Abstract

In a service system having plural servers and storage systems connected via a network, a storage library is installed on all servers on which applications are running. All applications can use the storage library to create new data, read data, and update data in different types of backend storage systems. In one embodiment, the storage library holds a list of all storage systems which can be used for storing data, and manages capabilities of the storage systems, such as data replication and data integrity checking. The storage library is arranged to create a write profile for each created data in order to indicate how the data is being managed. A storage system and the storage library are thus able to avoid performing unnecessary data management (for example, redundant replication) by checking whether such data management is or has been performed by the storage library or by the storage system by referring to the write profile.

Claims

1. A service system, comprising:

a plurality of servers arranged for applications to run thereon;

in each server, a storage library arranged such that the applications use the storage library to read/write data via a network to/from plural storage nodes of different types;

wherein each server is arranged to hold a storage list of storage nodes that store objects, and the storage library in each server is arranged to manage capabilities of the storage nodes, including object replication in response to a request by an application for the object replication or object integrity checking in response to a request by the application for the object integrity checking, and in the event the storage node targeted by the request lacks the capability to perform the requested object replication or object integrity checking;

wherein the storage library in each server is further arranged to create a write profile for each object created, the write profile including a storage node location for holding an object replica before replication of the object is performed, and information indicating whether an object replication is to be performed by the storage library or by a storage node having a location for holding the object replica;

wherein each storage node is arranged to hold a write profile of an object indicating whether the object is stored in a storage node holding the write profile; and

wherein in response to a request by an application for object replication, and if the write profile of the object specifies replication by the storage library, the storage library writes an object data together with its write profile to a primary storage node identified in the write profile as storing an original of the object, and writes the object data together with its write profile to all secondary storage nodes identified in its write profile as storage nodes in which replicas of the object data are stored.

2. The service system according to claim 1, wherein the secondary storage nodes are arranged to store object data replicas made by both the storage library and the storage node having the primary storage.

3. The service system according to claim 1, wherein the storage library in each server is arranged so that all applications requesting access to object data held in one or more of the storage nodes are able to access the object data without regard to type of storage node.

4. The service system according to claim 1, wherein the storage nodes include at least one node having no object management capability, at least one node having an object replication capability but no object integrity checking capability, and at least one node having both object replication capability and object integrity checking capability.

5. The service system according to claim 1, wherein object replication relationships among storage nodes are held in the storage list and updated by changing the storage list as storage nodes are added or removed, regardless of storage node type.

6. The service system according to claim 1, wherein the storage library determines, from an object write request received from an application, whether the storage library or the storage node as the primary storage node of the write request is to perform any replication of the object data.

7. The service system according to claim 6, wherein when the primary storage node is determined to perform a replication of the object data, the primary storage node reads the write profile of the object to be replicated, checks whether the next storage node as the secondary storage node is specified in the write profile, if the next storage node as the secondary storage node is specified in the write profile, the primary storage node does not perform the replication, but if the next storage node as the secondary storage node is not specified in the write profile, the primary storage node performs the replication to the next storage node.

8. A method for managing capabilities of a storage library and storage nodes in a service system, comprising the steps of:

creating a write profile of an object to be written in response to a write request;

adding a specification to the write profile that indicates whether the object is to be replicated by the storage library or by a primary storage node identified in the write profile and to which the object is to be written in response to the write request;

writing the object data and the write profile to the primary storage node identified in the write profile; and

replicating the object data and write profile to a secondary storage node identified in the write profile.

9. The method according to claim 8,

wherein the storage library is part of a service system having plural servers and plural storage nodes connected to the servers by a network, each server having one of said storage library;

and

wherein the step of creating a write profile includes steps of;

setting, by the storage library, an object name of the object in the write profile;

calculating, by the storage library, a hash value from the object name;

determining, by the storage library, the primary storage node that manages a hash space that includes the calculated hash value; and

setting an IP address of the primary storage node in the write profile.

10. The method according to claim 9, wherein the adding step comprises the steps of:

checking, by the storage library, whether a number of replicas specified in the write request to be created of the object is not zero;

if the number of replicas is not zero, checking the write profile, by the storage library, to determine whether the primary storage node is able to replicate data by itself;

if the primary storage node is able to replicate data by itself, checking the write profile, by the storage library, to determine whether a synchronous replication is required for replicating the object;

if a synchronous replication is required, determining, by the storage library, whether the primary storage node can perform a synchronous replication;

if the primary storage node can perform a synchronous replication, checking, by the storage library, whether replications by the storage node mean a request from an application by judging whether the number of replicas in the write request are greater than the number of replicas of the storage node;

if replications by the storage node are requested by the application, calling, by the storage library, a procedure by which the primary storage node replicates the object;

if a synchronous replication is not required, setting in the write profile, by the storage library, the number of replicas requested in the write request to be the number of replicas to be created;

if replications by the storage node are not requested by the application, setting in the write profile, by the storage library, the number of replicas requested in the write request minus the number of replicas to be performed by a storage node; and

calling, by the storage library, a procedure by which the storage library performs the replication.

11. The method according to claim 10, wherein the step of calling a procedure by which the storage library replicates the object comprises the steps of:

setting in a storage list, by the storage library, an IP address of the primary storage node specified in the write profile as a Next_Storage_Node variable indicating the next storage node which is a target of replication;

calling, by the storage library, a procedure by which the next storage node of the primary node is determined, and storing the next storage node of the primary storage node as the Next_Storage_Node variable;

checking, by the storage library, whether the next storage node is already specified in the write profile as a secondary storage node, which is a target of replication, and if the next storage node is already so specified, looking for another storage node as a next storage node;

if the next storage node is not already so specified, setting in the write profile, by the storage library, the next storage node as the secondary storage node;

checking, by the storage library, whether it has finished finding the same number of storage nodes as the number of replicas specified in the write profile, and if not, looking for another storage node;

if the storage library has finished finding the same number of storage nodes as the number of replicas specified in the write profile, specifying in the write profile, by the storage library, that replication is required to be performed by the storage library; and

setting in the write profile, by the storage library, the replication type field of the write profile to the same value as the replication type field of the write request.

12. The method according to claim 11, wherein the step of calling a procedure by which the next storage node of the primary node is determined comprises the steps of:

finding, by the storage library, a storage which has a replication relationship to a secondary storage node specified as the next storage node;

if such a storage node cannot be found, determining, by the storage library, a storage node having the next hash space to that of the storage node specified as the next storage node.

13. The method according to claim 8, further comprising the step of:

determining whether a data integrity check for the object data Is required by the storage library;

wherein the determining step comprises the steps of;

checking, by the storage library, whether the write request requires a data integrity check;

If the write request requires a data integrity check, checking the write profile, by the storage library, whether the primary storage node has a data integrity check capability; and

if the primary storage node does not have a data integrity check capability, performing, by the storage library, a data integrity check for the object data.

14. The method according to claim 8, wherein when a read request is received from an application, the storage library performs a method comprising the steps of:

determining whether load balancing of the storage nodes is required; and

determining whether a data integrity check of the read object data is required;

wherein the step of determining whether load balancing is required comprises the steps of;

determining, by the storage library, a storage node which has an original of the read object;

checking, by the storage library, whether load balancing is requested in the read request;

if load balancing is requested in the read request, finding, by the storage library, all storage nodes which have replicas by calling a procedure by which the next storage node of the primary node is determined;

if the number of replicas was not provided from the application, reading, by the storage library, the write profile to determine how many objects are replicated, but if the number of replicas cannot be read from the write profile, the storage library determines the storage nodes that have replicas from a prestored replication relationship;

selecting, by the storage library, one of the storage nodes having replicas, whose selection improves load balancing;

reading, by the storage library, the object requested in the read request together with its write profile from the determined storage node;

checking, by the storage library, whether the object read succeeded;

if the object read did not succeed, trying, by the storage library, another storage node which has a replica by again calling the procedure by which the next storage node of the primary node is determined;

if the object read succeeded, checking the write profile, by the storage library, whether the storage node performs data integrity checking;

if the storage node does not perform data integrity checking, the storage library calculates a data integrity value of the object data, and checks whether the calculated data integrity value is equal to a data integrity value specified in the write profile; and

if the calculated data integrity value is equal to the data integrity value specified in the write profile, the read object is transmitted to the requesting application.

15. A computer-readable medium containing computer-readable information which, when executed by a computer, causes the computer to perform a method for managing capabilities of a storage library and storage nodes in a service system comprising the steps of:

creating a write profile of an object to be written in response to a write request;

adding a specification to the write profile that indicates whether the object is to be replicated by the storage library or by a primary storage node identified in the write profile and to which the object is to be written in response to the write request;

writing the object data and the write profile to the primary storage node identified in the write profile; and

replicating the object data and write profile to a secondary storage node identified in the write profile.

16. The computer-readable medium according to claim 15,

wherein the storage library is part of a service system having plural servers and plural storage nodes connected to the servers by a network, each server having one of said storage library;

and

wherein the step of creating a write profile includes steps of;

setting, by the storage library, an object name of the object in the write profile;

calculating, by the storage library, a hash value from the object name;

determining, by the storage library, the primary storage node that manages a hash space that includes the calculated hash value; and

setting an IP address of the primary storage node in the write profile.

17. The computer-readable medium according to claim 16, wherein the adding step comprises the steps of:

checking, by the storage library, whether a number of replicas specified in the write request to be created of the object is not zero;

if the number of replicas is not zero, checking the write profile, by the storage library, to determine whether the primary storage node is able to replicate data by itself;

if the primary storage node is able to replicate data by itself, checking the write profile, by the storage library, to determine whether a synchronous replication is required for replicating the object;

if a synchronous replication is required, determining, by the storage library, whether the primary storage node can perform a synchronous replication;

if the primary storage node can perform a synchronous replication, checking, by the storage library, whether replications by the storage node mean a request from an application by judging whether the number of replicas in the write request are greater than the number of replicas of the storage node;

if replications by the storage node are requested by the application, calling, by the storage library, a procedure by which the primary storage node replicates the object;

if a synchronous replication is not required, setting in the write profile, by the storage library, the number of replicas requested in the write request to be the number of replicas to be created;

if replications by the storage node are not requested by the application, setting in the write profile, by the storage library, the number of replicas requested in the write request minus the number of replicas to be performed by a storage node; and

calling, by the storage library, a procedure by which the storage library performs the replication.

18. The computer-readable medium according to claim 17, wherein the step of calling a procedure by which the storage library replicates the object comprises the steps of:

setting in a storage list, by the storage library, an IP address of the primary storage node specified in the write profile as a Next_Storage_Node variable indicating the next storage node which is a target of replication;

calling, by the storage library, a procedure by which the next storage node of the primary node is determined, and storing the next storage node of the primary storage node as the Next_Storage_Node variable;

checking, by the storage library, whether the next storage node is already specified in the write profile as a secondary storage node, which is a target of replication, and if the next storage node is already so specified, looking for another storage node as a next storage node;

if the next storage node is not already so specified, setting in the write profile, by the storage library, the next storage node as the secondary storage node;

checking, by the storage library, whether it has finished finding the same number of storage nodes as the number of replicas specified in the write profile, and if not, looking for another storage node;

if the storage library has finished finding the same number of storage nodes as the number of replicas specified in the write profile, specifying in the write profile, by the storage library, that replication is required to be performed by the storage library; and

setting in the write profile, by the storage library, the replication type field of the write profile to the same value as the replication type field of the write request; and

wherein the step of calling a procedure by which the next storage node of the primary node is determined comprises the steps of;

finding, by the storage library, a storage node which has a replication relationship to a secondary storage node specified as the next storage node;

if such a storage node cannot be found, determining, by the storage library, a storage node having the next hash space to that of the storage node specified as the next storage node.

19. The computer-readable medium according to claim 15, further comprising the step of:

determining whether a data integrity check for the object data Is required by the storage library;

wherein the determining step comprises the steps of;

checking, by the storage library, whether the write request requires a data integrity check;

If the write request requires a data integrity check, checking the write profile, by the storage library, whether the primary storage node has a data integrity check capability; and

if the primary storage node does not have a data integrity check capability, performing, by the storage library, a data integrity check for the object data.

20. The computer-readable medium according to claim 15, wherein when a read request is received from an application, the storage library performs a method comprising the steps of:

determining whether load balancing of the storage nodes is required; and

determining whether a data integrity check of the read object data is required;

wherein the step of determining whether load balancing is required comprises the steps of;

determining, by the storage library, a storage node which has an original of the read object;

checking, by the storage library, whether load balancing is requested in the read request;

if load balancing is requested in the read request, finding, by the storage library, all storage nodes which have replicas by calling a procedure by which the next storage node of the primary node is determined;

if the number of replicas was not provided from the application, reading, by the storage library, the write profile to determine how many objects are replicated, but if the number of replicas cannot be read from the write profile, the storage library determines the storage nodes that have replicas from a prestored replication relationship;

selecting, by the storage library, one of the storage nodes having replicas, whose selection improves load balancing;

reading, by the storage library, the object requested in the read request together with its write profile from the determined storage node;

checking, by the storage library, whether the object read succeeded;

if the object read did not succeed, trying, by the storage library, another storage node which has a replica by again calling the procedure by which the next storage node of the primary node is determined;

if the object read succeeded, checking the write profile, by the storage library, whether the storage node performs data integrity checking;

if the storage node does not perform data integrity checking, the storage library calculates a data integrity value of the object data, and checks whether the calculated data integrity value is equal to a data integrity value specified in the write profile; and

if the calculated data integrity value is equal to the data integrity value specified in the write profile, the read object is transmitted to the requesting application.