US 2013/0085990 A1 · REPLICA PLACEMENT STRATEGY FOR DISTRIBUTED DATA PERSISTENCE

Abstract

Methods and systems are described that involve replica placement strategy for distributed systems. At the time of index creation, initial index replica placement is decided. The first index replica is placed locally on the current processing server node. This server node operates as a master host for the first replica. The other index replicas are placed on different hosts depending on the index type, split index or non-split index, and a set of criteria. For non-split indexes, the set of criteria includes: a number of recently created replicas per host, resource usage per host, and a total number of replicas per host. For split indexes, the set of criteria includes: a number of replicas of any split index part per host and a number of local first replicas of split index parts. If all criteria re equal, the first host in alphanumeric ordering receives a next replica.

Claims24

Dependent claims

A non-transitory computer-readable storage medium tangibly storing machine-readable instructions thereon, which when executed by the machine, cause the machine to perform operations comprising:

detecting creation of one or more of a non-split index and a split index, wherein the split index includes a plurality of split parts;
storing a first replica of a part of the split index on a local host in a plurality of hosts, wherein the first replica is a local replica and the local host is a master host for the local replica of the split index part;
determining a total number of replicas of any part of the split index that are stored on a first slave host;
when the total number of replicas stored on the first slave host is equal to the total number of replicas stored on a second slave host, determining a total number of local replicas stored on the first slave host and on the second slave host;
and
when the total number of local replicas stored on the first slave host is equal to the total number of local replicas stored on the second slave host, storing a second replica of the part of the split index on the first slave host.

3 more elements

A computer system, comprising:

a computer memory to store program code;
and
a processor to execute the program code to:
detect creation of one or more of a non-split index and a split index, wherein the split index includes a plurality Of split parts;
store a first replica of a part of the split index on a local host in a plurality of hosts, wherein the first replica is a local replica and the local host is a master host for the local replica of the split index part;
determine a total number of replicas of any part of the split index that are stored on a first slave host;
when the total number of replicas stored on the first slave host is equal to the total number of replicas stored on a second slave host, determine a total number of local replicas stored on the first slave host and on the second slave host;
and
when the total number of local replicas stored on the first slave host is equal to the total number of local replicas stored on the second slave host, store a second replica of the part of the split index on the first slave host.

6 more elements

Dependencies are inferred from the claim text.

Replica placement strategy for distributed data persistence

Life of this patent

Replica placement strategy for distributed data persistence

Life of this patent

Abstract

Claims24