#community-help

Benchmarks and Concurrent Users for Typesense

TLDR A inquired about the number of concurrent users Typesense can support. Jason suggested benchmarking for better accuracy, while Kishore Nallan mentioned about an upcoming feature. Jason and Kishore Nallan both recommended tools and gave references for performing benchmarks.

Powered by Struct AI

2

1

1

May 19, 2021 (33 months ago)
A
Photo of md5-98c72c3023867be0346b48ae4cb22001
A
04:27 AM
how many concurrent users searching can typesense support at 512 mb and 1gb?
Jason
Photo of md5-8813087cccc512313602b6d9f9ece19f
Jason
04:28 AM
It would totally depend on the type of queries and the size of the data that’s retrieved with every search
04:29
Jason
04:29 AM
So getting a representative set of queries and doing a benchmark is the best way to get an accurate picture
Kishore Nallan
Photo of md5-4e872368b2b2668460205b409e95c2ea
Kishore Nallan
04:29 AM
In the next version, we will be introducing some basic caching support that will pretty much greatly increase performance if your dataset does not change too often or you are okay in serving slightly stale results (e..g 60 seconds lag).

2

A
Photo of md5-98c72c3023867be0346b48ae4cb22001
A
04:30 AM
yes, my data doesn’t change that much
04:30
A
04:30 AM
i’m just trying to understand how many users it can support — any idea on how to go about the benchmark? Jason
Jason
Photo of md5-8813087cccc512313602b6d9f9ece19f
Jason
04:31 AM
04:32
Jason
04:32 AM
04:32
Jason
04:32 AM
It uses an OSS load testing framework called k6
Kishore Nallan
Photo of md5-4e872368b2b2668460205b409e95c2ea
Kishore Nallan
04:34 AM
Another quick option will be to use siege which allows you to load a set of URLs from a text file. So you can generate a set of query URLs and then ask siege to hit it: https://www.linode.com/docs/guides/load-testing-with-siege/

1

Jason
Photo of md5-8813087cccc512313602b6d9f9ece19f
Jason
04:34 AM
Also here are some benchmark numbers on different datasets to give you a rough ballpark idea of concurrency: https://github.com/typesense/typesense#benchmarks

1

Typesense

Lightning-fast, open source search engine for everyone | Knowledge Base powered by Struct.AI

Indexed 3015 threads (79% resolved)

Join Our Community

Similar Threads

Troubleshooting Typesense Document Import Error

Christopher had trouble importing 2.1M documents into Typesense due to memory errors. Jason clarified the system requirements, explaining the correlation between RAM and dataset size, and ways to tackle the issue. They both also discussed database-like query options.

3

30
10mo

Enhancing Vector Search Performance and Response Time using Multi-Search Feature

Bill faced performance issues with vector search using multi_search feature. Jason and Kishore Nallan suggested running models on a GPU and excluding large fields from the search. Through discussion, it was established that adding more CPUs and enabling server-side caching could enhance performance. The thread concluded with the user reaching a resolution.

3

140
1mo

Addressing Typesense Server Issues and Optimization Needs

Robert had an issue with a 'stuck' typesense server. Jason and Kishore Nallan gave advice on handling writes, configuration for high search volumes, and running multiple typesense instances. They also recommended monitoring CPU usage and updating the server version for bug fixes.

1

30
14mo

Understanding Indexing and Search-As-You-Type In Typesense

Steven had queries about indexing and search-as-you-type in Typesense. Jason clarified that bulk updates are faster and search-as-you-type is resource intensive but worth it. The discussion also included querying benchmarks and Typesense's drop_tokens_threshold parameter, with participation from bnfd.

2

13
29mo

Discussion on Performance and Scalability for Multiple Term Search

Bill asks the best way for multi-term searches in a recommendation system they developed. Kishore Nallan suggested using embeddings and remote embedder or storing and averaging vectors. Despite testing several suggested solutions, Bill continued to face performance issues, leading to unresolved discussions about scalability and recommendation system performance.

3

105
1w