Best Practices for Query Performance Testing and Tuning with Pinecone

This exclusive webinar explores the intricacies of testing and tuning Pinecone for optimal query performance and the best practices for performance testing of search applications.
In this webinar, you will understand the importance of performance testing and why it's the only reliable way to hit your target service level and appropriately size your Pinecone cluster. We will guide you through the basics of performance testing, including defining your test objectives, setting up a query performance test, and analyzing results. Additionally, we will discuss how to build candidate test indexes using ETL frameworks like Lucille and set up a query load testing tool like Gatling.
You will learn how and when to scale Pinecone vertically or horizontally to accommodate your performance targets. A live demo will include example performance tests for real-world scenarios like Semantic search and Hybrid search. The demo will show how to tune for query latency and throughput.
Take advantage of this unique opportunity to learn the best query performance testing and tuning practices with Pinecone. Register today and join us on May 11th at noon EST to unlock the full potential of your search application.
Featuring:
Brian Nauheimer - Cofounder, KMW Technology
Kyle Marcotte - Senior Search Engineer, KMW Technology

[@NavneetVermalivefree]

I have 2 questions:
1. What was the recall when you did the vector search only?
2. When you did the hybrid search what recall you got with Pinecone and Elastic Search?

[@sundarramanp3057]

@41.29 shows a good architecture diagram for doing hybrid search and document search storage. I have 2 questions about this architecture:
1. Pinecone itself supports hybrid search right? Why is elastic search also used along with pinecone for keyword/lexical search instead of pinecone only for both? Are there any pros of this approach?
2. Why are document chunks stored in elasticsearch db? Why not store it as a metadata {"text"} in pinecone vector db itself? Is it for memory cost reasons with pinecone or latency reasons that the actual document content is stored in elasticsearch db and only indexes of chunks are stored in pinecone?

[@BrunoSabadini7]

That's gold!