Overview
Experimenting and investigating CEL and the CEL-go engine
Benchmarking
Virtual Machine specs
goos: windows
goarch: amd64
pkg: github.com/nreisch/cel-go-tests
cpu: Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
Benchmarking Hot Path Compilation vs Evaluation
| Test Name | Number of Policy Expressions | Time (ns) per op | Bytes per op | Allocs |
|---|---|---|---|---|
| BenchmarkTest_Policy1_CompileEvaluate-16 | 1 | 61825 ns/op | 23416 B/op | 476 allocs/op |
| BenchmarkTest_Policy1_Evaluate-16 | 1 | 10301 ns/op | 6486 B/op | 60 allocs/op |
| BenchmarkTest_Policy100_CompileEvaluate-16 | 100 | 5825781 ns/op | 2343144 B/op | 47551 allocs/op |
| BenchmarkTest_Policy100_Evaluate-16 | 100 | 1097703 ns/op | 650685 B/op | 6062 allocs/op |
Benchmarking Arrays Evaluation
| Test Name | Number of Policy Expressions | Array Input Length | Time (ns) per op | Bytes per op | Allocs |
|---|---|---|---|---|---|
| BenchmarkTest_Policy1_Array1_Evaluate-16 | 1 | 1 | 14320 ns/op | 8243 B/op | 98 allocs/op |
| BenchmarkTest_Policy100_Array1_Evaluate-16 | 100 | 1 | 1687981 ns/op | 829979 B/op | 9936 allocs/op |
| BenchmarkTest_Policy100_Array10_Evaluate-16 | 100 | 10 | 1815824 ns/op | 889816 B/op | 11267 allocs/op |
| BenchmarkTest_Policy100_Array100_Evaluate-16 | 100 | 100 | 3079584 ns/op | 1436478 B/op | 29633 allocs/op |
| BenchmarkTest_Policy1000_Array1_Evaluate-16 | 1000 | 1 | 15934536 ns/op | 8745264 B/op | 109681 allocs/op |
| BenchmarkTest_Policy1000_Array10_Evaluate-16 | 1000 | 10 | 17464611 ns/op | 9376829 B/op | 123757 allocs/op |
| BenchmarkTest_Policy1000_Array100_Evaluate-16 | 1000 | 100 | 28511772 ns/op | 15166177 B/op | 314868 allocs/op |
| BenchmarkTest_Policy1000_Array1000_Evaluate-16 | 1000 | 1000 | 145696300 ns/op | 84891340 B/op | 2213605 allocs/op |
References
- https://github.com/google/cel-go/
- https://github.com/google/cel-spec/
- https://github.com/google/cel-policy-templates-go
- https://codelabs.developers.google.com/codelabs/cel-go
Commands
go test -count=1 -bench=. -benchmem