Elasticsearch Suggester: 10 Times Performance Degradation from v.2.3.3 to v.5.1.2

Am I new noticing this?

I have huge autosuggester index created in version 2.3.3, about 235,000 records total, for City Autosuggest feature, taken from reallife database (including misspellings), and it is 10x faster than 4,000 cities (loaded from Geonames.org) "autosuggest" field of Elasticsearch v.5.1.2

And both results are far slower than initially reported in this blog: https://www.elastic.co/blog/you-complete-me

  • to recall, it was reported 0,5 ms response time for the whole Wikipedia (which is over five millions articles in English)!!!

I use "hey" from https://github.com/rakyll/hey to run tests, 50 parallel threads, 10,000 total queries, and my dataset is much smaller; and v.2..3 with 235,000 records is 10x faster than v.5.1.2!

Is it anything already well known? Thanks,

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