OPERATION METHOD AND APPARATUS FOR DATA STORAGE SYSTEM

• US 20140089259A1
• Filed: 11/29/2013
• Published: 03/27/2014
• Est. Priority Date: 06/01/2011
• Status: Active Grant

• Current Assignee: Huawei Technologies Co., Ltd. (Huawei Investment & Holding Co., Ltd.)
• Original Assignee: Huawei Technologies Co., Ltd. (Huawei Investment & Holding Co., Ltd.)
• Inventors: CHENG, Shi
• Granted Patent: US 9,639,588 B2
• Time in Patent Office: Days
• US Class Current: 707/612
• CPC Class Codes: G06F 16/27 Replication, distribution or synchronisation of data between databases or within a distributed database system; Distributed database system architectures therefor
• 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

An operation method and apparatus for a data storage system are provided, which relate to the field of data storage and enhance efficiency of a read operation on distributed stored data. The method includes: receiving, by a data node, a data operation message, where the data operation message carries a data operation type and a data object of an operation; querying a state parameter of the data object, where the state parameter is allocated to the data node by a management node, and is used to indicate an update state of the data object, and the update state includes a synchronous state, a semi-synchronous state and an asynchronous state; and if the state parameter is the synchronous state and the operation type is a read operation, reading the data object stored by the data node. Embodiments of the invention are used to operate data in a distributed data storage system.

Claims

1. An operation method for a data storage system, comprising:

receiving, by a data node, a data operation message, wherein the data operation message carries a data operation type and a data object of an operation;

querying a state parameter of the data object, wherein the state parameter is allocated to the data node by a management node, and is used to indicate an update state of the data object, and the state parameter comprises a synchronous state, a semi-synchronous state and an asynchronous state; and

if the state parameter is the synchronous state and the operation type is a read operation, reading the data object stored by the data node.

2. The method according to claim 1, further comprising:

if the state parameter is the semi-synchronous state or the asynchronous state and the operation type is the read operation, querying state parameters being the synchronous state from a node state table saved by the data node and acquiring other data nodes with the state parameters being the synchronous state, wherein the node state table records data saved by each data node and a state parameter corresponding to the saved data; and

sending a read operation message to one data node among the other data nodes, wherein the read operation message comprises the data object, so that the other data nodes read the data object after receiving the read operation message.

3. The method according to claim 1, wherein the querying the state parameter of the data object comprises:

when a management state parameter of the data node is a normal state, querying the state parameter of the data object; and

when the management state parameter of the data node is a connection interruption state, ending a current data operation.

4. The method according to claim 2, after the data node sends the read operation message to the one data node among the other data nodes, further comprising:

if a management state exception message returned by the one data node because a management state parameter is a connection interruption state is received within a preset first waiting time period, sending the read operation message to another data node among the other data nodes.

5. The method according to claim 4, wherein the updating the state parameter of the data object of the one data node recorded in the node state table of the data node comprises:

acquiring a state parameter of the data object of the one data node from the update state exception message; and

changing the state parameter of the data object of the one data node recorded in the node state table to the state parameter of the data object acquired from the update state exception message.

6. The method according to claim 4, further comprising:

periodically sending, by the data node, the timeout record table of the data node to the management node, so that the management node modifies state parameters of all the data of the data node reaching a timeout threshold to the asynchronous state and recording the asynchronous state into a management node state table;

receiving, by the data node, a management node state table sent by the management node; and

updating the node state table of the data node according to the management node state table.

7. The method according to claim 6, wherein if the total number of data nodes which save data whose state parameters are the synchronous state and the semi-synchronous state is lower than a preset node number threshold, the management node selects a data node with the least timeout records from data nodes which save data whose state parameters are the asynchronous state, and changes a state parameter of data stored by the data node with the least timeout records to the semi-synchronous state.

8. The method according to claim 6, wherein the updating, by the data node, the node state table of the data node according to the management node state table comprises:

acquiring a state parameter of the data saved by each data node from the management node state table; and

when the acquired state parameter of the data saved by each data node is different from a state parameter in the node state table of the data node, updating the state parameter in the node state table of the data node to the acquired state parameter of the data saved by each data node.

9. The method according to claim 7, after the updating the node state table of the data node according to the management node state table, further comprising:

if a state parameter of partial data of the data node is changed to the semi-synchronous state after the node state table of the data node is updated, copying the partial data from another data node which stores the partial data, the state parameter of which is the synchronous state; and

after copying of the partial data is completed, sending a synchronization completion message to the management node, so that the management node changes the state parameter of the partial data of the data node to the synchronous state in the management node state table after receiving the synchronization completion message; and

the management node dynamically adjusts a state parameter of data in the management node state table, so that the total number of data nodes which store any kind of data whose state parameters are the synchronous state and data nodes which store any kind of data whose state parameters are the semi-synchronous state does not exceed a node number threshold.

10. The method according to claim 6, further comprising:

when the number of consecutive connection failures of a management connection request initiated by the data node to the management node reaches a preset connection failure counting threshold, setting, by the data node, a management state parameter of the data node itself to a connection interruption state.

11. The method according to claim 1, further comprising:

if the state parameter is the synchronous state and the operation type is a write operation, writing the data object stored by the data node;

if the state parameter is the semi-synchronous state or the asynchronous state and the operation type is the write operation, querying state parameters being the synchronous state from a node state table saved by the data node and acquiring other data nodes with the state parameters being the synchronous state;

after the other data nodes with the state parameters being the synchronous state are acquired, sending a first write operation message to one data node among the other data nodes with the state parameters being the synchronous state, wherein the first write operation message comprises the data object, so that other data nodes write the data object after receiving the first write operation message; and

after the data node or the one data node among the other data nodes with the state parameters being the synchronous state completes writing of the data object, sending a second write operation message to all other data nodes which save the data object, the state parameter of which is the synchronous, semi-synchronous or asynchronous state, so that all the other data nodes write the data object.

22. The method according to claim 2, after the data node sends the read operation message to the one data node among the other data nodes, further comprising:

if an update state exception message returned by the one data node because the state parameter of the data object is not the synchronous state is received within a preset first waiting time period, updating a state parameter of the data object of the one data node recorded in the node state table of the data node; and

after updating of the state parameter is completed, sending the operation message to another data node among the other data nodes.

23. The method according to claim 2, after the data node sends the read operation message to the one data node among the other data nodes, further comprising:

if a response of the one data node is not received within a first waiting time period, increasing the number of timeouts of the one data node by 1 in a timeout record table stored by the data node.

12. An operation apparatus for a data storage system, comprising:

an operation message receiving unit, configured to receive a data operation message, wherein the data operation message carries a data operation type and a data object of an operation;

a state parameter querying unit, configured to query a state parameter of the data object, wherein the data parameter is allocated to a data node by a management node, and is used to indicate an update state of the data object, and the state parameter comprises a synchronous state, a semi-synchronous state and an asynchronous state; and

a data reading unit, configured to, when the state parameter is the synchronous state and the operation type is a read operation, read the data object stored by the data node.

13. The apparatus according to claim 12, further comprising:

a first node querying unit, configured to, when the state parameter is the semi-synchronous state or the asynchronous state and the operation type is the read operation, query state parameters being the synchronous state from a node state table saved by the data node and acquire other data nodes with the state parameters being the synchronous state, wherein the node state table records data saved by each data node and a state parameter corresponding to the saved data; and

a read operation message sending unit, configured to send a read operation message to one data node among the other data nodes, wherein the read operation message comprises the data object, so that the other data nodes read the data object after receiving the read operation message.

14. The apparatus according to claim 12, wherein the state parameter querying unit is further configured to query the state parameter of the data object when a management state parameter of the data node is a normal state, and end a current data operation when the management state parameter of the data node is a connection interruption state.

15. The apparatus according to claim 13, after the data node sends the read operation message to the one data node among the other data nodes, further comprising:

a second read operation sending unit, configured to, after a management state exception message returned by the one data node because a management state parameter is a connection interruption state is received within a preset first waiting time period, send the read operation message to another data node among the other data nodes;

a state parameter updating unit, configured to, when an update state exception message returned by the one data node because the state parameter of the data object is not the synchronous state is received within the preset first waiting time period, update a state parameter of the data object of the one data node recorded in the node state table of the data node, and after updating of the state parameter is completed, send the read operation message to another data node among the other data nodes; and

a timeout state updating unit, configured to, when a response of the one data node is not received within the first waiting time period, increase the number of timeouts of the one data node by 1 in a timeout record table stored by the data node.

16. The apparatus according to claim 15, further comprising:

a management connection request unit, configured to initiate a management connection request to the management node by taking a preset time period as a cycle;

a timeout record sending unit, configured to send a timeout record table of the data node to the management node after the management connection request is received by the management node, so that the management node modifies state parameters of all the data of the data node reaching a timeout threshold to the asynchronous state according to the timeout record table and record the asynchronous state into a management node state table;

a management state receiving unit, configured to receive a management node state table sent by the management node; and

a node state table updating unit, configured to update the node state table of the data node according to the management node state table.

17. The apparatus according to claim 16, wherein the node state table updating unit comprises:

a first state parameter acquiring module, configured to acquire a state parameter of the data object of the one data node from the update state exception message; and

a first state parameter updating module, configured to change the state parameter of the data object recorded in the node state table to the state parameter of the data object acquired from the update state exception message.

18. The apparatus according to claim 17, wherein the node state table updating unit further comprises:

a second state parameter acquiring module, configured to acquire a state parameter of the data saved by each data node from the management node state table; and

a second state parameter updating module, configured to, when the acquired state parameter of the data saved by each data node is different from a state parameter in the node state table of the data node, update the state parameter in the node state table of the data node to the acquired state parameter of the data saved by each data node.

19. The apparatus according to claim 17, after updating the node state table of the data node according to the management node state table, further comprising:

a synchronous data copying unit, configured to, when a state parameter of partial data of the data node is changed to the semi-synchronous state, copy the partial data from another data node which stores the partial data, the state parameter of which is the synchronous state; and

a synchronization completion message sending unit, configured to send a synchronization completion message to the management node after copying of the partial data is completed, so that the management node changes the state parameter of the partial data of the data node to the synchronous state in the management node state table after receiving the synchronization completion message; and

the management node dynamically adjusts a state parameter of data in the management node state table, so that the total number of the data nodes which store any kind of data whose state parameters are the synchronous state and the data nodes which store any kind of data whose state parameters are the semi-synchronous state does not exceed a preset node number threshold.

20. The apparatus according to claim 17, further comprising:

a connection interruption state setting unit, configured to, when the number of consecutive connection failures of a management connection request initiated by the data node to the management node reaches a preset connection failure counting threshold, set the management state parameter of the data node itself to a connection interruption state by the data node.

21. The apparatus according to claim 12, further comprising:

a data writing unit, configured to, when the state parameter is the synchronous state and the operation type is a write operation, write the data object stored by the data node;

a second node querying unit, configured to, when the state parameter is the semi-synchronous state or the asynchronous state and the operation type is the write operation, query state parameters being the synchronous state from a node state table saved by the data node and acquire other data nodes with the state parameters being the synchronous state;

a first write operation sending unit, configured to send a first write operation message to one data node among the other data nodes with the state parameters being the synchronous state after the other data nodes with the state parameters being the synchronous state are acquired, wherein the first write operation message comprises the data object, so that the other data nodes write the data object after receiving the first write operation message; and

a second write operation sending unit, configured to, after the data node or the one data node among the other data nodes with the state parameters being the synchronous state completes writing of the data object, send a second write operation message to all other data nodes which save the data object, the state parameter of which is the synchronous state, the semi-synchronous state or the asynchronous state, so that all the other data nodes write the data object.