During the last 3 months, I worked on the Quickwit search engine to adapt it for AWS Lambda runtime. I wrote a blog post to announce the beta release [0] and two tutorials to quickstart:

- Indexing and searching 20 million logs entries dataset [1].

- Implementing an E2E use case where an application generates data, uploads it in batches to S3. Users can then search through an HTTP API authenticated with an API Key [2].

On the performance side, I observed on a 20 million log dataset that search is sub-second, and analytics queries take from 1 to 4 second.

On the cost side, obviously the compute using Lambda is fairly expensive, but the huge win is that it scales to 0. Few other solutions do. For instance OpenSearch Serverless has a fixed cost around $700/months. On the search side, thanks to indexing, queries are usually way cheaper than scanning services like CloudWatch (at least 50x cheaper).

I would love to get the feedback from the community and see how I can push this further.

[0]: Blog post https://quickwit.io/blog/quickwit-lambda-beta

[1]: 20 million log dataset search tutorial https://quickwit.io/docs/get-started/tutorials/tutorial-aws-...

[2]: E2E use case tutorial https://quickwit.io/docs/guides/e2e-serverless-aws-lambda