DISTRIBUTED STORAGE SYSTEM AND DISTRIBUTED STORAGE METHOD

• US 20140019683A1
• Filed: 06/16/2011
• Published: 01/16/2014
• Est. Priority Date: 03/16/2011
• Status: Active Grant

• Current Assignee: NEC Corporation
• Original Assignee: NEC Corporation
• Inventors: Ishikawa, Kenichiro
• Granted Patent: US 9,317,221 B2
• Time in Patent Office: Days
• US Class Current: 711/114
• CPC Class Codes: G06F 16/1827 Management specifically adapted to NAS management of storage area networks [SAN] G06F3/067
• CPC Class Codes: G06F 3/061 Improving I/O performance
• CPC Class Codes: G06F 3/0659 Command handling arrangements, e.g. command buffers, queues, command scheduling
• CPC Class Codes: G06F 3/067 Distributed or networked storage systems, e.g. storage area networks [SAN], network attached storage [NAS]

Abstract

A distributed storage system comprises a plurality of data storage nodes that store a plurality of object data items in a distributed manner. The plurality of object data items are respectively characterized by a data coordinate and identified by a data identifier. The plurality of data storage nodes are respectively assigned a node coordinate in a space identical to that of the data coordinate. After selecting as first data storage nodes from among the plurality of data storage nodes a plurality of data storage nodes each assigned a node coordinate having a relatively short distance from a data coordinate characterizing each of the plurality of object data items, any of the object data item is respectively stored in a data storage node determined based on a data identifier that identifies the object data item from among the plurality of first data storage nodes.

Claims

1. A distributed storage system, comprising:

a plurality of data storage nodes that store a plurality of object data items in a distributed manner, wherein

the plurality of object data items are respectively characterized by a data coordinate and uniquely identified by a data identifier,

the plurality of data storage nodes are respectively assigned a node coordinate in a space identical to that of the data coordinate, and

after selecting as first data storage nodes from among the plurality of data storage nodes a plurality of data storage nodes each assigned a node coordinate having a relatively short distance from a data coordinate characterizing each of the plurality of object data items, any of the object data items is respectively stored in a data storage node determined based on a data identifier that identifies the object data item from among the plurality of first data storage nodes.

2. The distributed storage system according to claim 1, wherein

data coordinates characterizing each of the plurality of object data items and node coordinates assigned to each of the plurality of data storage nodes are coordinates in a Euclidean space having as coordinate axes a plurality of evaluation axes with which the plurality of object data items are quantitatively evaluated.

3. The distributed storage system according to claim 1, wherein

a plurality of data storage nodes each assigned a node coordinate within a prescribed distance from an object data item are selected as first data storage nodes from among the plurality of data storage nodes.

4. The distributed storage system according to claim 1, wherein

a prescribed number of data storage nodes, each assigned a node coordinate having a relatively short distance from an object data item, are selected in order from shortest distance as first data storage nodes from among the plurality of data storage nodes.

5. The distributed storage system according to claim 1, wherein

a plurality of pseudo random numbers generated using a data identifier that identifies an object data item as a seed are respectively assigned to the plurality of first data storage nodes, and a data storage node having a maximum value that is obtained by multiplying with an assigned pseudo random number a weight calculated according to a distance between a data coordinate of the object data item and a node coordinate assigned to each of the plurality of first data storage nodes, is extracted from among the plurality of first data storage nodes, and the extracted data storage node is determined as a data storage node to store the object data item concerned.

6. The distributed storage system according to claim 5, wherein

the weight increases as a distance between a data coordinate characterizing each of the plurality of object data items and a node coordinate assigned to each of the plurality of data storage nodes decreases.

7. The distributed storage system according to claim 1, wherein

each of the plurality of data storage nodes is assigned a node coordinate(s) according to a distribution of the data coordinates characterizing the plurality of object data items.

8. The distributed storage system according to claim 1, further comprising:

a client node that writes an object data item to the plurality of data storage nodes, and reads an object data item stored in the plurality of data storage nodes.

9. The distributed storage system according to claim 8, wherein

the client node selects as first data storage nodes from among the plurality of data storage nodes a plurality of data storage nodes each assigned a node coordinate that has a relatively short distance from a data coordinate characterizing an object data item, and writes the object data item in a data storage node determined based on a data identifier of the object data item from among the plurality of first data storage nodes.

10. The distributed storage system according to claim 8, wherein

the client node selects as first data storage nodes from among the plurality of data storage nodes a plurality of data storage nodes each assigned a node coordinate that has a relatively short distance from a data coordinate characterizing an object data item, and sends a command to read the object data item to the selected plurality of first data storage nodes.

11. The distributed storage system according to claim 8, wherein

the client node selects as first data storage nodes from among the plurality of data storage nodes a plurality of data storage nodes each assigned a node coordinate that has a relatively short distance from a data coordinate characterizing an object data item, and sends a command to read the object data item to a data storage node determined based on a data identifier that identifies the object data item from among the plurality of first data storage nodes.

12. A distributed storage method, comprising:

storing in a distributed manner a plurality of object data items each characterized by a data coordinate and uniquely identified by a data identifier;

assigning a node coordinate in a space identical to that of the data coordinate to each of the plurality of data storage nodes;

selecting as first data storage nodes from among the plurality of data storage nodes a plurality of data storage nodes each assigned a node coordinate having a relatively short distance from a data coordinate characterizing each of the plurality of the object data items; and

storing the object data item in a data storage node determined based on a data identifier that identifies the object data item from among the plurality of first data storage nodes.

13. The distributed storage method according to claim 12, wherein

the data coordinates characterizing each of the plurality of object data items and node coordinates assigned to each of the plurality of data storage nodes are coordinates in a Euclidean space having as coordinate axes a plurality of evaluation axes with which the plurality of object data items are quantitatively evaluated.

14. The distributed storage method according to claim 12, comprising:

selecting as first data storage nodes from among the plurality of data storage nodes a plurality of data storage nodes each assigned a node coordinate within a prescribed distance from an object data item.

15. The distributed storage method according to claim 12, comprising:

selecting a prescribed number of data storage nodes, each assigned a node coordinate having a relatively short distance from an object data item, in order from shortest distance as first data storage nodes from among the plurality of data storage nodes.

16. The distributed storage method according to claim 12, comprising:

assigning a plurality of pseudo random numbers, generated using a data identifier of an object data item as a seed, respectively to the plurality of first data storage nodes;

extracting from among the plurality of first data storage nodes a data storage node having a maximum value that is obtained by multiplying with an assigned pseudo random number a weight calculated according to a distance between a data coordinate of the object data item and a node coordinate assigned to each of the plurality of first data storage nodes; and

determining the extracted data storage node as a data storage node to store the object data item.

17. The distributed storage method according to claim 12, comprising:

assigning node coordinates to each of the plurality of data storage nodes in accordance with a distribution of the data coordinates characterizing the plurality of object data items.

18. The distributed storage method according to claim 16, wherein

the weight increases as a distance between a data coordinate characterizing each of the plurality of object data items and a node coordinate assigned to each of the plurality of data storage nodes decreases.