How many concurrent writes does a Elasticsearch single node count as normal?


(willdx) #1

How many concurrent writes does a Elasticsearch single node count as normal?

Environment : ELK 5
Machine performance: 4 core /16GB/SSD hard disk

Problem 1: Elasticsearch single node write speed (concurrent quantity) how many is normal?

I wrote the log to redis first, and then transferred it to Elasticsearch by logstash. The amount of concurrency was about 1500-2000 seconds, and I felt that the amount of concurrent Elasticsearch support of a single node was not very large. Is that normal?

Problem 2: in addition, if I want to improve the performance of writing to Elastic, can I only use clustering?

For example, if I want to achieve 4000/s concurrency, can I just add machines to the Elastic cluster?


(Christian Dahlqvist) #2

The indexing throughput a single node can sustain can depend on a number of factors, e.g.:

  • Size and complexity of document
  • Complexity of mappings
  • Size of bulk requests
  • How document IDs are assigned
  • Hardware specification (CPU cores and performance of storage)
  • Number of shards being indexed into

Without this information it is difficult to judge your achieved throughput.

Forming a cluster by adding additional nodes will generally improve indexing throughput, but if you at the same time add a replica shard in order to gain resilience you have also doubled the amount of work the cluster need to do for each document.


(system) #3

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.