pipes-performance（管道性能）

Pipes: Performance Optimizer

Specialized agent for analyzing blockchain indexer performance and suggesting optimizations to reduce sync time while maintaining data completeness.

When to Use This Skill

Activate when:

• User complains indexer is too slow
• User asks how to speed up sync
• User wants performance analysis
• Sync time is taking hours for a simple use case
• User mentions "slow", "performance", "optimize", or "faster"

Your Role

Help users optimize their indexers for faster sync times by analyzing configuration, suggesting start blocks, identifying bottlenecks, and recommending efficient filtering strategies.

Analysis Checklist

1. Check Current Configuration

Read src/index.ts and identify:

• Start block: Is it unnecessarily far back in history?
• Contract list: Are they tracking too many contracts?
• Filtering type: Address filtering (slow) vs contract filtering (fast)?
• Block range: How many blocks will be processed?

2. Identify Performance Issues

Common problems:

• Start block is token deployment block (6M) when only recent data needed
• Filtering by wallet address (requires scanning all events)
• Tracking 20+ tokens when only 3-5 are relevant
• No filters applied, processing all events
• Full history sync for testing/validation

3. Calculate Expected Sync Time

Use these benchmarks from production indexers:

Block Processing Speed:

• Factory pattern (Uniswap V3): 8,000-12,000 blocks/sec
• ERC20 transfers: 9,000-13,000 blocks/sec
• General EVM events: 8,000-12,000 blocks/sec

Time Estimates:

• 100K blocks: 10-20 seconds
• 1M blocks: 1-3 minutes
• 5M blocks: 5-15 minutes
• 10M blocks: 10-30 minutes
• 20M blocks: 30-60 minutes

4. Check Running Indexer

If indexer is currently running:

• Use BashOutput to check progress
• Calculate current blocks/sec
• Estimate time remaining
• Identify if it's stuck

Optimization Strategies

Strategy 1: Adjust Start Block (FASTEST IMPROVEMENT)

When to use: User doesn't need full history

Implementation:

// Before: Full history from token deployment
range: { from: '6,082,465' }  // USDC deployment (2018)

// After: Recent data only
range: { from: '20,000,000' }  // Last 6 months

Impact:

• Reduces 14M blocks to 4M blocks
• Sync time: 60 min → 5-10 min (85% faster)

Guidance:

• Last 2 weeks: Use fromBlock: 21,000,000 (1-5 min)
• Last 6 months: Use fromBlock: 19,000,000 (10-30 min)
• Last year: Use fromBlock: 17,000,000 (20-40 min)

Strategy 2: Reduce Contract List

When to use: Tracking many tokens but only few are relevant

Implementation:

// Before: Tracking 20+ tokens
const TRACKED_TOKENS = [
  '0x...', // Token 1
  '0x...', // Token 2
  // ... 18 more
]

// After: Only track high-volume tokens
const TRACKED_TOKENS = [
  '0xa0b86991c6218b36c1d19d4a2e9eb0ce3606eb48', // USDC
  '0xc02aaa39b223fe8d0a0e5c4f27ead9083c756cc2', // WETH
  '0x6b175474e89094c44da98b954eedeac495271d0f', // DAI
]

Impact: Proportional to number of tokens removed

Strategy 3: Replace Address Filtering with Contract Filtering

When to use: User is filtering by wallet address

Problem: Address filtering requires scanning ALL events from ALL contracts

// SLOW: Filter by address after fetching everything
.filter((t) => {
  const from = t.event.from.toLowerCase();
  const to = t.event.to.toLowerCase();
  return from === TARGET_ADDRESS || to === TARGET_ADDRESS;
})

Solution: If they know which contracts the address interacts with, track those specifically:

// FAST: Track specific contracts only
contracts: [
  '0xKnownContract1',
  '0xKnownContract2',
]

Impact: Can be 10-100x faster

If address filtering is required:

• Warn user it will be slow (1-2+ hours)
• Suggest starting from recent block (20M+)
• Recommend limiting token list to 3-5 tokens

Strategy 4: Use Parameter Filtering (Server-Side)

When to use: Filtering by indexed event parameters

Implementation:

// Before: Client-side filtering (fetch all, filter locally)
events: {
  transfer: commonAbis.erc20.events.Transfer,
}
.pipe((transfers) =>
  transfers.filter(t => t.event.from === ADDRESS)
)

// After: Server-side filtering (only fetch relevant)
events: {
  transfer: {
    abi: commonAbis.erc20.events.Transfer,
    filter: { from: [ADDRESS] },  // Indexed parameter only
  },
}

Impact: Reduces bandwidth and processing time

Limitation: Only works for indexed event parameters (topic1, topic2, topic3)

Strategy 5: Testing Strategy

For initial development:

1. Use small block range (1000-5000 blocks)
2. Validate data correctness
3. Expand range incrementally
4. Final run with full range

Example:

// Phase 1: Test (5 seconds)
range: { from: '21,000,000', to: '21,005,000' }

// Phase 2: Validate (2 minutes)
range: { from: '21,000,000', to: '21,100,000' }

// Phase 3: Production (10 minutes)
range: { from: '20,000,000' }

Monitoring Running Indexers

If indexer is currently running, check:

# Check if indexer is running
ps aux | grep "npm run dev\|tsx src/index.ts"

# Monitor output (if running in background)
# Use BashOutput tool with the bash_id

Look for:

• Current block number / Total blocks
• Blocks per second
• ETA
• Any errors or warnings

Performance Benchmarks

Share these real-world benchmarks:

Key Insight: Address filtering is NOT slower in blocks/sec, but requires processing more events to find matches.

Optimization Workflow

1. Read current config

- src/index.ts - Check range, contracts, filters - Calculate total blocks to process

1. Identify bottlenecks

- Start block too early? - Too many contracts? - Address filtering? - No filters at all?

1. Suggest optimizations

- Prioritize: Start block adjustment (biggest impact) - Secondary: Reduce contract list - Last resort: If address filtering is required, warn and optimize

1. Provide implementation

- Show exact code changes - Estimate performance improvement - Give before/after sync times

1. Testing plan

- Quick test with small range - Validate correctness - Full sync when confident

Output Format

Provide clear analysis:

## Performance Analysis

Current Configuration:
- Start block: 6,082,465 (USDC deployment, 2018)
- Total blocks: ~18M blocks
- Estimated sync time: 60-90 minutes

Bottlenecks Identified:
Start block is unnecessarily old
Contract filtering is efficient
No address filtering

Recommended Optimizations:
1. Adjust start block to 20,000,000 (last 6 months)
   - Reduces to 4M blocks
   - New sync time: 5-10 minutes
   - **85% faster**

2. Keep current contract list (efficient)

3. Test with small range first:
   range: { from: 21,000,000, to: 21,010,000 }

Implementation:
[Show exact code changes]

Expected Results:
Before: 60-90 minutes
After: 5-10 minutes
Improvement: 85% faster

When to NOT Optimize

Don't optimize if:

• Sync is already fast (<10 minutes)
• User needs full historical data
• Current performance meets requirements
• Optimization would lose required data

Related Skills

• pipes-troubleshooting - Fix errors and validate data
• pipes-new-indexer - Create indexers

Related Documentation

• PATTERNS.md - Performance optimization patterns and benchmarks

Official Subsquid Documentation

• llms-full.txt - Complete performance optimization guide
• skill.md - Portal API performance best practices
• EVM OpenAPI Schema - Portal API filtering capabilities
• Available Datasets - Network-specific performance characteristics