2GB Week RICE Framework Feedback
The output of this initial focus on reducing the memory footprint of GitLab will be issues describing approaches to memory reduction. We will use the RICE Framework for scoring and prioritization.
Topics to keep in mind while scoring
- (Impact) Estimated memory reduction - e.g. reduce by 100MB
- Effort Level - difficult, moderate, easy
- (Effort) Time estimate - perhaps using t-shirt sizing? days or weeks?
- Impact - what portion of the product will this impact
- (Effort) Isolation level - how isolated is the change?
- Others
Issues to review for RICE scoring (link to database scoring example)
- Sidekiq cluster should preload before forking - gitlab-org/gitlab#215317
- Map object allocations to features - gitlab-org/gitlab#289844
- Optimise GC and
jemalloc
of all other processes other than GitLab Rails - gitlab-org/gitlab#289842 (closed) - Optimise GC and
jemalloc
of GitLab - gitlab-org/gitlab#289838 (closed) - Ensure that all gems use
frozen_string_literal: true
- gitlab-org/gitlab#289839 - Drop gitlab-exporter - gitlab-org/gitlab#288043 (closed)
- Find mechanism to dynamically load gems - gitlab-org/gitlab#288044 (closed)
- Run GC compact before forking into Puma workers - gitlab-org/gitlab#288042 (closed)
- Extend RubyVM with frozen string literal deduplication - gitlab-org/gitlab#289840 (closed)
Edited by Craig Gomes