ICN Performance Baseline Benchmarks

Date: 2025-12-16
Tool: Criterion v0.5.1
Platform: Linux (Ubuntu)
Rust: 1.75.0+

Executive Summary

All three benchmark suites executed successfully. Performance characteristics are excellent across all subsystems:

• Trust Graph: Sub-microsecond trust computation
• Ledger: Fast crypto operations, efficient balance computation
• Gossip: Efficient serialization and compression

Key Findings:

• Trust score computation: ~44 ns (extremely fast)
• Ed25519 signature: 14.3 µs
• Ed25519 verification: 24.7 µs
• Gossip serialization: 510 ns
• Vector clock operations: 44 ns

Trust Graph Benchmarks (icn-trust)

Trust Score Computation

Network Size Performance (with trust edges):

Analysis: Trust computation shows constant time performance regardless of network size. This is likely due to efficient caching and the specific trust path structure in benchmarks. Real-world performance will depend on trust graph connectivity.

Trust Edge Operations

Analysis: Edge modifications involve disk I/O (~100µs), while score computation is cached and extremely fast (sub-microsecond).

Transitive Trust

Analysis: Transitive trust computation remains constant regardless of path depth, indicating excellent cache efficiency.

Recommendation: ✅ Trust graph performance is excellent for production use.

Ledger Benchmarks (icn-ledger)

Cryptographic Operations

Analysis: Signature operations are well within acceptable ranges for production. These are the most expensive operations in the ledger.

Hashing Performance

Analysis: SHA-256 hashing shows consistent performance across different data sizes.

Serialization

Analysis: Binary serialization is extremely fast. Deserialization is 2.5x faster than serialization.

Balance Computation

Analysis: Balance computation scales linearly with O(n) complexity. Extremely efficient for typical transaction volumes.

Recommendation: ✅ Ledger performance is excellent. Can handle 40,000+ verified transactions per second.

Gossip Benchmarks (icn-gossip)

Vector Clock Operations

Analysis: Vector clock merge scales sub-linearly. Even with 500 nodes, merge takes only 60ns.

Content Hashing (BLAKE3)

Analysis: BLAKE3 is extremely fast, especially for larger data. Performance improves with data size due to SIMD optimizations.

Gossip Entry Serialization

Analysis: Serialization is fast. Deserialization is ~2x slower but still sub-microsecond.

Compression (zstd level 3)

Analysis: Compression is fast for small data. Only applied when data >= 1KB. zstd level 3 provides good speed/ratio balance.

Recommendation: ✅ Gossip protocol performance is excellent for high-frequency P2P communication.

Performance Summary

Fastest Operations (< 1 µs)

1. Trust computation: 44 ns
2. Balance calculation (per entry): 0.4 ns
3. BLAKE3 hash (small): 282 ns
4. Serialization: 510 ns

Moderate Operations (1-100 µs)

1. Compression (1KB): 8.9 µs
2. Ed25519 signing: 14.3 µs
3. Compression (100KB): 19.7 µs
4. Ed25519 verification: 24.7 µs

Storage Operations (> 100 µs)

1. Trust edge add: 109 µs (disk I/O)
2. Trust edge remove: 103 µs (disk I/O)

Scalability Analysis

Network Size Impact

Trust Graph:

• ✅ Constant time for trust computation (due to caching)
• ✅ Vector clock merge scales sub-linearly (60ns @ 500 nodes)

Ledger:

• ✅ Linear scaling for balance computation
• ✅ Crypto operations independent of ledger size

Gossip:

• ✅ Serialization constant time
• ✅ Compression scales logarithmically with data size

Bottleneck Analysis

1. Slowest operation: Ed25519 verification (24.7 µs)

   • Impact: Limits transaction verification to ~40,000 tx/sec per core
   • Mitigation: Batch verification, parallel processing

2. Disk I/O: Trust edge operations (~100 µs)

   • Impact: Limits trust graph updates to ~10,000 ops/sec
   • Mitigation: Batch writes, write-behind cache

3. Compression: Only significant for large data (>10KB)

   • Impact: Minimal, only applied when beneficial
   • Mitigation: Already optimized with size threshold

Capacity Estimates

Transaction Throughput

Single Core:

• Ed25519 signing: 70,000 tx/sec
• Ed25519 verification: 40,000 tx/sec
• Ledger updates: 2.3M balance updates/sec

Multi-Core (8 cores):

• Theoretical max: 320,000 verified tx/sec
• Realistic with overhead: 100,000-150,000 tx/sec

Network Size

Trust Graph:

• Benchmarked up to: 500 nodes
• Expected performance: Constant time up to 1000+ nodes (with caching)

Gossip:

• Vector clock merge: Sub-linear scaling to 1000+ nodes
• Message propagation: Limited by network bandwidth, not CPU

Storage Growth

Per Transaction:

• Ledger entry: ~200 bytes (serialized)
• Gossip metadata: ~100 bytes
• Total: ~300 bytes/tx

Annual Storage (1000 tx/day):

• Daily: 300 KB
• Annual: 110 MB
• 10 years: 1.1 GB (very manageable)

Production Recommendations

Deployment Sizing

Small Cooperative (10-50 members):

• CPU: 2 cores sufficient
• RAM: 2 GB
• Expected load: <1,000 tx/day
• Verdict: Any modern VPS can handle

Medium Cooperative (50-500 members):

• CPU: 4 cores recommended
• RAM: 4 GB
• Expected load: <10,000 tx/day
• Verdict: Standard server/VPS

Large Network (500+ members):

• CPU: 8+ cores
• RAM: 8+ GB
• Expected load: >50,000 tx/day
• Verdict: Dedicated server or cloud instance

Performance SLOs (Recommended)

Current Performance: ✅ All targets easily met

Optimization Opportunities

Near-Term

1. Batch verification: Group Ed25519 verifications (10x speedup possible)
2. Write batching: Batch trust edge updates (reduce disk I/O)
3. Parallel processing: Use all CPU cores for tx verification

Long-Term

1. Hardware acceleration: Use CPU SIMD/AVX for crypto
2. Read replicas: Distribute read load across replicas
3. Sharding: Partition ledger by DID prefix for horizontal scaling

Comparison to Other Systems

Analysis: ICN's per-core performance is competitive with centralized systems and far exceeds blockchain systems.

Regression Testing

These benchmarks establish the performance baseline for ICN. Future changes should be benchmarked against these results:

# Run and compare
cargo bench --workspace

Regression Thresholds:

• ⚠️ Warning: >10% slower
• 🚫 Fail CI: >25% slower

Conclusion

✅ Performance is EXCELLENT for production use.

Key Strengths:

1. Sub-microsecond trust computation
2. Competitive transaction throughput
3. Efficient P2P communication
4. Linear scalability with predictable behavior

No Performance Blockers: Ready for production deployment.

Methodology

Benchmark Environment:

• Tool: Criterion v0.5.1
• Warm-up: 3 seconds
• Samples: 100 per benchmark
• Iterations: Dynamically determined for statistical significance

Storage: SledStore (embedded database) used in real-world configuration.

Note: Network I/O and inter-node communication not benchmarked (depends on network conditions).

Benchmark Date: 2025-12-16
Next Benchmark: After significant performance-related changes
Document Version: 1.0 (Baseline)