Method for Providing Access to Data Items from a Distributed Storage System

• US 20120254342A1
• Filed: 09/27/2011
• Published: 10/04/2012
• Est. Priority Date: 09/28/2010
• Status: Abandoned Application

• Current Assignee: Metaswitch Networks Limited (Microsoft Corporation)
• Original Assignee: Metaswitch Networks Limited (Microsoft Corporation)
• Inventors: Evans, Michael Jeffrey
• Application Number:
  US13/246,781
• Publication Number:
  US 20120254342A1
• Time in Patent Office: Days
• US Class Current: 709/214
• CPC Class Codes: G06F 11/186 Passive fault masking when reading multiple copies of the same data
• CPC Class Codes: G06F 11/2064 while ensuring consistency
• CPC Class Codes: G06F 2201/82 Solving problems relating to consistency
• 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]
• CPC Class Codes: H04L 69/40 for recovering from a failure of a protocol instance or entity, e.g. service redundancy protocols, protocol state redundancy or protocol service redirection management of faults, events, alarms or notifications in data switching networks H04L41/06

Abstract

A method for providing access to data items from a distributed storage system is provided. Each data item is replicated across a plurality of storage nodes. Data items are read from the distributed storage system by selecting between a first reading mode, comprising attempting to read the data item from a set of the storage nodes to check for data item consistency across at least a quorum of the set of nodes, and a second reading mode, comprising reading the data item from at least one of the storage nodes. The reading mode is selected according to at least one detected characteristic of system status of the distributed storage system. The second reading mode is selected when the detected characteristic indicates a higher likelihood of data item consistency, and the first reading mode is selected when the detected characteristic indicates a lower likelihood of data item consistency.

Claims

1. A method for providing access to data items from a distributed storage system having multiple storage nodes connected via a data communications network, each data item being replicated across a plurality of the storage nodes, the method comprising:

receiving a request to retrieve the data item at a given node;

selecting, according to at least one detected characteristic of system status of the distributed storage system, a first or second reading mode to use in the retrieval of the data item, the first reading mode comprising attempting to read the data item from a set of storage nodes to check for data item consistency across at least a quorum of the set of storage nodes, the second reading mode comprising reading the data item from at least one of the storage nodes, said selecting the first or second reading mode comprising;

selecting the first reading mode when the detected characteristic of system status indicates a lower likelihood of data item consistency, and

selecting the second reading mode when the detected characteristic of system status indicates a higher likelihood of data item consistency;

reading the data item from the distributed storage system using the selected reading mode; and

transmitting the data item, read by the selected reading mode, in response to the request to retrieve the data item.

2. The method according to claim 1, wherein the first reading mode comprises reading the data item from a first number of the plurality of storage nodes, wherein the second reading mode comprises reading the data item from a second number of the plurality of storage nodes, and wherein said second number is less than said first number.

3. The method according to claim 1, wherein the second reading mode comprises reading the data item from only one of the plurality of storage nodes.

4. The method according to claim 1, wherein the selection of the reading mode is performed at the given node in response to the receipt of the request to retrieve the data item, in accordance with a current system status indicated by the detected characteristic of system status.

5. The method according to claim 1, wherein the quorum comprises more than half of the set of storage nodes.

6. The method according to claim 4, further comprising updating the data item stored at one or more of the plurality of storage nodes using the data item read from the quorum of the plurality of storage nodes.

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

requesting messages from a plurality of the set of storage nodes, indicating agreement to a write command for the data item;

detecting an agreement issue characteristic indicating that the data item write command has not yet been agreed by set of storage nodes;

including the agreement issue characteristic in the detected characteristic of system status of the distributed storage system.

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

monitoring for an indication that a locally-stored data item may be out of date at the given node;

detecting an out-of-date characteristic indicating that the locally-stored data item may not be a latest version of the data item;

including said out-of-date characteristic in the detected characteristic of system status of the distributed storage system.

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

allocating a current leader node for the set of storage nodes;

detecting a leadership issue characteristic indicating that the given node is not the current leader for the set of storage nodes;

including the leadership issue characteristic in the detected characteristic of system status of the distributed storage system.

10. The method according to claim 9, further comprising, at the given node:

assuming responsibility for controlling updates to the data item stored at each member of the set of storage nodes in response to receipt at the given node of an acknowledgement message from at least one of the other storage nodes in the set of storage nodes; and

detecting the leadership issue characteristic from a leadership configuration message received at the given node, which leadership configuration message indicates that a subsequent change in the responsibility for controlling updates has occurred.

11. The method according to claim 8, wherein said monitoring includes monitoring whether said given node is in a recently-failed state.

12. The method according claim 8, wherein said monitoring includes monitoring whether said given node has been unable to make contact with one or more other storage nodes in said set of storage nodes.

13. The method according to claim 8, wherein said monitoring includes monitoring whether an instance of updating relating to the data item has been completed unexpectedly.

14. A method for providing backup access to data items from a storage node of a distributed storage system having multiple storage nodes connected via a data communications network, each data item being replicated across a plurality of the storage nodes, the method comprising:

detecting that the storage node has been disconnected from other storage nodes in a set of storage nodes across which data items have been replicated;

querying one or more members of the set of storage nodes to identify a current leader node which is responsible for coordinating updates to the set of storage nodes;

determining one or more data items in relation to which the storage node has missed an update to the set of storage nodes; and

in response to said determining, querying one or more of the set of storage nodes to update a locally-stored data item to ensure consistency of the locally-stored data item with at least a quorum of the set of storage nodes.

15. The method according to claim 14, wherein said determining one or more data items in relation to which the storage node has missed an update to the set of storage nodes comprises identifying one or more data items in relation to which the leader node has not confirmed completion of a replication action.

16. A method for providing backup access to data items from a storage node of a distributed storage system having multiple storage nodes connected via a data communications network, each data item being replicated across a plurality of the storage nodes, the method comprising:

detecting that the storage node has failed and restarted;

querying one or more members of a set of storage nodes to identify a current leader node which is responsible for coordinating updates to the set of storage nodes;

determining an identifier of a point in a sequence of distributed storage system actions being coordinated by the current leader node, and participating in read and/or write actions instructed by the leader node which are beyond the point in the sequence.

17. A computer program product comprising a non-transitory computer-readable storage medium having computer readable instructions stored thereon, the computer-readable instructions being executable by a computerized device to cause the computerized device to perform a method for providing access to data items from a distributed storage system having multiple storage nodes connected via a data communications network, each data item being replicated across a plurality of the storage nodes, the method comprising:

receiving a request to retrieve the data item at a given node;

selecting, according to at least one detected characteristic of system status of the distributed storage system, a first or second reading mode to use in the retrieval of the data item, the first reading mode including attempting to read the data item from a set of storage nodes to check for data item consistency across at least a quorum of the set of storage nodes, the second reading mode including reading the data item from at least one of the storage nodes, said selecting the first or second reading mode including;

selecting the first reading mode when the detected characteristic of system status indicates a lower likelihood of data item consistency, and

selecting the second reading mode when the detected characteristic of system status indicates a higher likelihood of data item consistency;

reading the data item from the distributed storage system using the selected reading mode; and

transmitting the data item, read by the selected reading mode, in response to the request to retrieve the data item.