Skip to content

Conversation

@vmmon
Copy link

@vmmon vmmon commented Nov 13, 2025

Currently each write is done synchronously. Thus Rocksdb performs a fdatasync() system-call for any transaction. This is not only slow but also leads to significant write amplification on the SSD which wears down it quite fast.

By using asynchronous writes we let the kernel flush the modified buffers periodically. This batching lead to an 8x throughput improvement in the soak-test.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant