-
Notifications
You must be signed in to change notification settings - Fork 2.1k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
compact: Redesign compaction planning process for cost efficiency and determinism. #3405
Comments
Potentially blocker as bigger compactions would require more resources: #3406 |
Hello 👋 Looks like there was no activity on this issue for the last two months. |
Still valid |
Hello 👋 Looks like there was no activity on this issue for the last two months. |
Very valid |
Hello 👋 Looks like there was no activity on this issue for the last two months. |
Closing for now as promised, let us know if you need this to be reopened! 🤗 |
Closing for now as promised, let us know if you need this to be reopened! 🤗 |
Still valid |
Hello 👋 Looks like there was no activity on this issue for the last two months. |
Hey, dear @yeya24 and @bwplotka. The proposal that I attached to #3406, tries to solve this issue as well. https://drive.google.com/file/d/1oZd3ENSZ7v2hONNf4pDL3yXr_0eHcwsV/view?usp=sharing Looking forward your feedback. |
Sweet, thanks, adding to my Todo list. |
Hello 👋 Looks like there was no activity on this issue for the last two months. |
Related discussion: https://matrix.to/#/!WaUKIfoqfiyWQhenET:matrix.org/$16044993991172247rFDCZ:matrix.org?via=matrix.org&via=hoffie.info
Problems of the Current Algorithm
The current algorithm is not very efficient for the cost (e.g network bandwidth). For example, if you have blocks laid out as 2h 2h 2h 2h 2h 2h 2h 2h 2h 2h 2h 2h 2h (301 of those in total), our current planning with comp range e.g [2h, 8h, 2d, 25d] will pick the oldest 4x2h to compact 8h and so on, then same for 2d only to at then do a single block of 25d. This means 88 compactions instead of just SINGLE compaction 300x2h blocks into single 25d. Vertical compactions are totally excluded as separate compactions so it's even worse.
It's currently (without locking mechanism) hard to run multiple compactors on a single stream. It would be nice to apply some determinism to allow such compactor sharding across a single stream. This is however lower priority goal.
This is related to #3390
The text was updated successfully, but these errors were encountered: