Performance Optimizations Tracker

[dapplion]

📋 Milestone/Epic Tracker

This tracker is to monitor outstanding issues and PRs relating to performance optimizations for Lodestar:

🟥 High Priority

• I/O lag issue in under-resourced machine Extreme I/O lag on overloaded nodes #4002
• Use faster Ed25519 impl for gossip crypto #3465
  • Impact: 5% total CPU
  • Idea: Use native gossip crypto
• gossipsub decode performance #2457
  • Impact: 10% total CPU
  • Idea: Faster gossip deserialization
• altair processAttestations performance issue #3080
  • Impact: 90% of processBlock
  • Idea: Optimize processAttestations
• Ban star exports #3116
  • Idea: Reduce export cost
• protons performance issue #3000
  • Impact: 5% of total CPU
  • protons lib performance issue
• Node stability improvements #2488
• Make State Cache Pruning Smarter #2154

🟧 Medium Priority

• Refactor gossip validation to return instead of throwing #3402
  • Idea: Optimize att gossip validation
• Verify signatures while processing on initial sync #2425
  • Idea: Initial sync verify sigs in parallel
• Partially TreeBacked block #3415
  • Idea: Only TreeBacked states
• Review regen state cache strategy #2846
  • Idea: Optimize regen cost
• Use optional chaining to optimize logging #3411
  • Idea: Optimize logger formatting
• Call produceAttestationData once for all validators #3098
  • Idea: Call produceAttestationData once
• processEth1Data performance issue #3082
  • Idea: Optimize processEth1Data
• Sync Committee attendance performance issue #3443

🟨 Low Priority

• Group attestations by target checkpoint #2877
  • Idea: Optimize att gossip validation
• Batch level db ops #3439
  • Idea: Batch level db ops
• Replace .map instances with for loops in Repository class (abstractRepository.ts) #2749
  • Idea: Use raw for loops
• Cache indexed attestations #3414
  • Idea: Cache indexed attestations
• Improve forkchoice updateHead #2747
  • Impact: 1% of total CPU
  • Idea: Improve forkchoice updateHead
• sync hashBlocks performance issue #2857
  • Idea: Sync hashBlocks performance issue
• Review RAND seeding on BLS batch verify #2527
  • Idea: Cheaper RAND seeding
• unshuffleList performance #2352
  • Idea: unshuffleList performance
• Lodestar tree hashing performance #2206
  • Idea: Improve hashing performance
• Consider minimizing GC runs if possible #2050
  • Idea: Minimize Garbage Collector GC cost

[philknows]

Not a scalable tracking solution for issues. Closing.