Performance Benchmark Report - vLLM Semantic Router

Generated: 2025-12-04 16:30:00 UTC

Git Commit: 816dbec26397

Git Branch: perf_test

Go Version: go1.24.1

32

Total Benchmarks

1

Regressions

8

Improvements

23

No Change

⚠️ WARNING: Performance regressions detected! Review the detailed results below.

🔍 Detailed Results

Benchmark	Metric	Baseline	Current	Change	Status
BenchmarkClassifyBatch_Size1	ns/op	10,245,678	10,123,456	-1.19%	OK
	P95 Latency	10.50ms	10.12ms	-3.62%
	Throughput	97.60 qps	98.78 qps	+1.21%
BenchmarkClassifyBatch_Size10	ns/op	52,345,678	51,234,567	-2.12%	🚀 IMPROVED
BenchmarkEvaluateDecisions_Complex	ns/op	456,789	512,345	+12.16%	⚠️ REGRESSION
	P95 Latency	0.46ms	0.52ms	+13.04%
	Throughput	2,189 qps	1,952 qps	-10.83%
BenchmarkCacheSearch_1000Entries	ns/op	3,456,789	3,389,012	-1.96%	🚀 IMPROVED
BenchmarkCacheConcurrency_50	ns/op	789,012	756,234	-4.16%	🚀 IMPROVED
	Throughput	1,267 qps	1,322 qps	+4.34%

📈 Performance Trends

📊 Interactive charts would appear here

Showing latency trends, throughput over time, and component comparisons

🔴 Regressions (Action Required)

Benchmark	Issue	Impact	Recommendation
BenchmarkEvaluateDecisions_Complex	P95 latency +13.04% Throughput -10.83%	Complex decision scenarios slowed significantly	Profile with `make perf-profile-cpu` Investigate rule matching optimization

Significant Improvements

Cache Concurrency: +4.34% throughput improvement under high load
Classification Batch Processing: Consistent 1-2% improvements across all batch sizes
Request Processing: 2.43% faster header/body handling