Index Partitioning
Index Partitioning improves the indexing and query performance by dividing a large index of documents across multiple partitions stored in one or more nodes.
The benefits include:
-
Scale out horizontally as the index size increases.
-
Increased indexing throughput by parallelising the indexing workload across partitions.
-
Reduction of query latency for large, faceted queries; since partitions are scanned in parallel.
| Updating the index partition count results in the complete rebuild of the index. So, please choose the index partition count after sizing the cluster for the business usecase. |
Index Partitions
The Index Partitions interface provides a section to enter the number of partitions the index is to be split into:
The default option for this setting is 1. Note that this number represents the number of primary partitions for an index, and the primary partitions are evenly distributed across all the nodes in the cluster where the search service is running. All the vbuckets within the couchbase bucket are evenly mapped to the index partitions configured here.
| Primary partitions represents the active partitions that serves queries. Replica partitions don’t serve queries and they solely aid the High Availability. |
The index partitions information is saved in its JSON definition, which can be previewed in the Index Definition Preview panel, at the right-hand side.
"planParams": {
"numReplicas": 0,
"indexPartitions": 1
},