Policy-based storage structure distribution

• US 9,367,600 B2
• Filed: 11/25/2015
• Issued: 06/14/2016
• Est. Priority Date: 04/16/2009
• Status: Active Grant

• Current Assignee: Cloud Software Group
• Original Assignee: TIBCO Software Incorporated (Cloud Software Group)
• Inventors: Moyne, Jean-Noel, Subramani, Suresh, Kim, Tom M.
• Primary Examiner(s): Jacob, Ajith
• Application Number:
  US14/952,884
• Publication Number:
  US 20160078117A1
• Time in Patent Office: 202 Days
• Field of Search: 707/661
• US Class Current: 1/1
• CPC Class Codes: G06F 16/178 Techniques for file synchronisation in file systems
• CPC Class Codes: G06F 16/2272 Management thereof
• CPC Class Codes: G06F 16/24554 Unary operations; Data partitioning operations
• CPC Class Codes: G06F 16/275 Synchronous replication

Abstract

Policy-based storage and retrieval combined with a distribution algorithm results in automatic and even distribution of policy-based storage structures across a set of nodes and dynamic, automated homing or ownership of policy-based storage structures. Large numbers of policy-based storage structures may be distributed without manual administration, allowing for rapid creation and destruction of storage structures. The overall load may be distributed and balanced across the server pool. Multiple entries having the same key value in a database- or table-like structure allow for distribution of policy-based storage and retrieval by key value and for queue semantics to be utilized for microqueues in the large database- or table-like structure.

Claims

1. A method of removing a leaving node from a collection of computing nodes of a distributed storage system, the method comprising:

the leaving node signaling to all other nodes in the collection of computing nodes that it is leaving the collection;

all the other nodes reading information from the leaving node;

all the other nodes synchronizing the information from the leaving node prior to the leaving node leaving the collection;

the leaving node leaving the collection of computing nodes, and

all the other nodes taking ownership of data associated with key values previously owned by the leaving node once the leaving node leaves the collection,

wherein all the other nodes reading information comprises all the other nodes receiving copies of the data associated with the key values owned by the leaving node, the date associated with the key values being the data whose ownership will be transferred from the leaving node to the other nodes once the leaving node leaves the collection.

2. The method of claim 1, further comprising replicating key values previously replicated on the leaving node.

3. The method of claim 1, wherein reading information from the leaving node further comprises calculating, using a consistent hashing algorithm, a distribution of key values over the collection of computing nodes excluding the leaving node.

4. A method for replicating data over a collection of computing nodes of a distributed storage system, the method comprising:

calculating a primary node for storing a policy-based storage structure associated with a key value;

calculating a secondary node for storing a copy of the policy-based storage structure associated with the key value; and

replicating the policy-based storage structure stored on the primary node and storing the copy of the policy-based storage structure on the secondary node,

wherein ownership of the policy-based storage structure is transferred from the primary node to the secondary node once the primary node leaves the collection.

5. The method of claim 4, wherein the calculating the primary node and the calculating the secondary node comprises using a distribution algorithm having monotone behavior.

6. The method of claim 4, wherein the calculating the primary node and the calculating the secondary node comprises using a consistent hashing algorithm.

7. The method of claim 6, wherein the primary node comprises the node closest to a mapped key value associated with the policy-based storage structure on a consistent hashing scale, and wherein the secondary node comprises the next closest node to the mapped key value associated with the policy-based storage structure on the consistent hashing scale.

8. A method of adding a node to a collection of computing nodes of a distributed storage system, the method comprising:

joining a new node to the collection of computing nodes as a leech node;

receiving, at the leech node, information from all other active nodes in the collection of computing nodes;

synchronizing the leech node based on the received information; and

thereafter, promoting the leech node to an active node in the collection of computing nodes;

wherein the joining the new node as a leech node comprises the leech node obtaining a key space from all of the other active nodes in the collection of computing nodes;

wherein prior to promoting the leech node to an active node, the leech node is not an active node, and

wherein the receiving information comprises receiving copies of key values from all other active nodes, the key values being key values whose ownership will be transferred from the other active nodes to the leech node once the leech node once it becomes an active node.

9. The method of claim 8, wherein the joining the new node as a leech node comprises the leech node interacting with the other active nodes in the collection of computing nodes.

10. The method of claim 8, wherein the receiving information comprises reconciling data between the leech node and the other active nodes in the collection of computing nodes.

11. The method of claim 8, wherein the receiving information comprises the leech node monitoring updated key values.

12. The method of claim 8, wherein receiving information further comprises calculating, using a consistent hashing algorithm, a distribution of key values over the collection of computing nodes including the new node.