Tutorial | Rule Engine | GRL | GRL JSON | RETE Algorithm | Functions | FAQ | Benchmark
For any library Benchmarking is very much required in terms of evaluating the performance and helps us to improve the performance better.
I have benchmarked two things:
- Loading 100 and 1000 rules into KnowledgeBase
- Executing a fact against rules against 100 and 1000 rules
All the tests will run for multiple times to see how it is performing by changing the value of N
N is b.N where b is an instance of *testing.B
refer 100_rules.grl and 1000_rules.grl files under examples/benchmark directory
> go test -bench=. -benchmem
goos: darwin
goarch: amd64
Number of Cores - 6
Ram - 16 GB 2400 MHz DDR4
pkg: github.com/hyperjumptech/grule-rule-engine/examples/benchmarkResult:
Load 100 and 1000 Rules into Grule rule Engine
Benchmark_Grule_Load_Rules/100_rules-12 67137 17387 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/100_rules#01-12 67485 17447 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/100_rules#02-12 67332 17408 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/100_rules#03-12 67992 17436 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/100_rules#04-12 68170 17420 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/100_rules#05-12 67777 17645 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/100_rules#06-12 65100 17431 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/100_rules#07-12 67396 17396 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/100_rules#08-12 68132 17458 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/100_rules#09-12 67881 17399 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/100_rules#10-12 67216 17523 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/1000_rules-12 66828 17823 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/1000_rules#01-12 69122 17581 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/1000_rules#02-12 67815 17425 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/1000_rules#03-12 67405 19681 ns/op 5070 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/1000_rules#04-12 48511 21222 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/1000_rules#05-12 67779 18999 ns/op 5070 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/1000_rules#06-12 56694 17691 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/1000_rules#07-12 69086 17641 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/1000_rules#08-12 51638 19401 ns/op 5070 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/1000_rules#09-12 58940 20498 ns/op 5071 B/op 118 allocs/op
Benchmark_Grule_Load_Rules/1000_rules#10-12 67411 19487 ns/op 5071 B/op 118 allocs/op
To load 100 rules into knowledgeBase it took 17523 ns/op (took the highest value) that is equal to ~0.017523ms and (5071 B/op) 0.005071MB memory per operation
To load 1000 rules into knowledgeBase it took 21222 ns/op (took the highest value) that is equal to ~0.021222ms and (5071 B/op) 0.005071MB memory per operation
Result:
Executing a fact against 100 and 1000 rules
Benchmark_Grule_Execution_Engine/100_rules-12 2055945 574 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/100_rules#01-12 2048078 570 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/100_rules#02-12 2086953 572 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/100_rules#03-12 2094231 571 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/100_rules#04-12 2078065 576 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/100_rules#05-12 2028356 642 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/100_rules#06-12 2002248 628 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/100_rules#07-12 1850121 703 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/100_rules#08-12 1761343 585 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/100_rules#09-12 2080953 594 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/100_rules#10-12 2082880 573 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/1000_rules-12 2082183 575 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/1000_rules#01-12 2098585 568 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/1000_rules#02-12 2090640 570 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/1000_rules#03-12 2109938 587 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/1000_rules#04-12 2045216 576 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/1000_rules#05-12 2092534 575 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/1000_rules#06-12 1994415 579 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/1000_rules#07-12 2098788 599 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/1000_rules#08-12 2092808 573 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/1000_rules#09-12 2085716 609 ns/op 512 B/op 9 allocs/op
Benchmark_Grule_Execution_Engine/1000_rules#10-12 1864302 576 ns/op 512 B/op 9 allocs/opTo execute a fact against 100 rules, Grule Engine took ~703 ns/op (took the highest value as base) that is hardly ~0.000703ms and (512 B/op) 0.000512MB memory per operation which is pretty fast.
To execute a fact against 1000 rules, Grule Engine took ~587 ns/op (took the highest value as base) that is hardly ~0.000587ms and (512 B/op) 0.000512MB memory per operation which is also pretty fast.