Skip to content

Latest commit

 

History

History
13 lines (9 loc) · 799 Bytes

PerformanceTuning.md

File metadata and controls

13 lines (9 loc) · 799 Bytes

Performance Tuning

  • Start with a configuration of the cluster and its nodes: data, master, and client nodes
  • Guesstimate usage pattern: lots of searches, vs. lots of indexing, vs. lots of aggregation
  • OS Configuration: number of open file descriptors shall be set to, at least, 32K-64K
  • OS Configuration: disable memory swapping
  • Mapping and data-modelling selection: _all, _source, and multi-fields mapping operations are increasing the load on the cluster
  • Search vs. Filter - do you need to compute a score based on the 'status: active' in your query?
  • Use _bulk Api for indexing whenever possible
  • Number of replicas and number of shards: availability and scaling-out vs. speed of writes dilemma