Elasticsearch Performance Tuning
Searches that used to return fast now crawl, aggregations spike CPU, or data nodes are constantly hot. This is Elasticsearch performance tuning — profiling the actual queries and cluster behavior, not generic checklists.
What I Can Help With
Tune Elasticsearch for your specific query and data patterns.
Typical Problems
Relevant Experience
Tuned Elasticsearch search performance for Niftycent (niftycent.com) — a US price intelligence engine. Profiled slow aggregations using the profile API, found fielddata pressure from a misconfigured text field, and restructured the mapping to use keyword + search_as_you_type.
What Is Included
How It Works
You share cluster stats, hot threads, slow search log and profile API output for a few slow queries. I identify the cause and recommend tuning steps. Typically one focused day.
Starting from €499
Profiling and a written set of tuning recommendations. Implementation is quoted separately.
Frequently Asked Questions
What causes high CPU on data nodes?
Usually expensive aggregations, poorly mapped text fields, fielddata usage, or too many shards per node. The profile API tells us exactly which one it is for your queries.
Do I need to reindex to tune performance?
Some fixes need a new index (mapping changes), most don't (query, aggregation, cache and refresh tuning). I list which is which so you can choose.
Related Services
Elasticsearch Optimization
Optimize Elasticsearch queries, mappings, indices and cluster configuration.
Custom Elasticsearch Development
Implement Elasticsearch for search, analytics and indexing.
Elasticsearch Migration & Upgrade
Migrate to or upgrade Elasticsearch with minimal downtime.
Backend Performance Audit
Find and prioritize the real bottlenecks across your full backend stack.
Systems I've Built
Cenoskop / Levnobot
Crawler monitoring millions of products. Processing 30M products daily and comparing them.
Built and maintain the full crawling pipeline — from scraping through Elasticsearch indexing to price comparison.
Niftycent
Price intelligence engine for the US market. Processing large-scale product data.
Built the data pipeline and crawling infrastructure for large-scale US product monitoring.
Not sure what's wrong?
Write what's happening and paste anything relevant — errors, logs, metrics. I'll have a look and tell you what I think.