Fulltext search has been optimized for better performance

Fulltext search has been optimized for better performance

In this release, search performance of different types of indices like single-tier (e.g. items only index) or multi-tier (e.g. 3-tier index having product-variant-item) has been further optimized. These optimizations also help to keep Elasticsearch clusters extremely cost-efficient by reducing the CPU and memory needed for them to scale.
By comparing performance between the initial 10.0 release and this release (10.0 HotFix 1), there is a performance improvement by up to factor 5 on an economical hardware setup for the Elasticsearch environment.
The following table illustrates the search response time examples tested for this release. Here are the parameters of the test setup:
  • Elasticsearch server is having 1 node of 8 CPUs and 32 GB RAM
  • 3-tier search index example having a data volume of 4.8 million records (500k Products, 700k Variants, 3.5m Items)
    • Products, variants and items, all had complex data including attributes
  • The index held language-specific data as well
Results from the test scenario above:
Operation
10.0
10.0 HotFix 1
Improvement
Subsequent searches
30 sec
3 sec
10x
Entity radio button changes in facets
40 sec
10 sec
4x
Normal facet selection
20 sec
4 sec
5x
Random text search
10 sec
5 sec
2x
Attribute facet selection
15 sec
4 sec
4x
It is mandatory to re-build the complete indices by using the simplified templates and creating export profile for the indices. Further details are available in the Migration Guide under the Web Search Migration chapter.

0 COMMENTS

We’d like to hear from you!