Mastering OpenSearch Benchmark: Advanced Techniques for Power Users
summary
Shows how OpenSearch Benchmark actually gets extended: register a Python generate_query_with_random_status parameter source to emit queries that pick HTTP statuses 200 304 404 at random, use Jinja2 templating for workload variables, and call raw REST requests when an API is unsupported. Demonstrates concurrent traffic via the parallel block, time‑based schedules for multi‑day longevity runs, telemetry (node stats, JFR) sampled per second into a metric store, and the benchmark compare flow that diffs benchmark-matrix, benchmark-results and test-executions to spot regressions.
AI-written summary · about
chapters (12)
- 0:00Welcome
- 1:04OpenSearch Benchmark basics
- 5:17Custom workloads with Ginger
- 7:23Dynamic queries with parameters
- 14:28Raw requests for unsupported APIs
- 15:53Telemetry devices
- 21:13Custom workloads for production
- 25:01Simulating concurrent traffic
- 27:30Longevity tests with schedules
- 30:35Compare test results
- 33:43Key takeaways
- 37:52Q&A
related talks
28mSuper Search with OpenSearch and Python
20mMeasuring Data with OpenSearch to Better Serve the Community
28mRunning OpenSearch at Scale in High-Traffic Gaming Systems
39mOpenSearch Helps You to Adopt a Friend
23mBeyond Slow Logs: Implementing Advanced Query Insights in OpenSearch
17mMastering the Query Insights Dashboard in OpenSearch
14mVector Search Benchmark[eting]
26mMeasuring Search Engine Quality using Spark and Python