This is a test suite for benchmarking various Go serialization methods.
- encoding/gob
- encoding/json
- github.com/alecthomas/binary
- github.com/davecgh/go-xdr/xdr
- github.com/Sereal/Sereal/Go/sereal
- github.com/ugorji/go/codec
- gopkg.in/vmihailenco/msgpack.v2
- github.com/youtube/vitess/go/bson (using the bsongen code generator)
- labix.org/v2/mgo/bson
- github.com/philhofer/msgp (code generator for msgpack)
- github.com/golang/protobuf (generated code)
- github.com/gogo/protobuf (generated code, optimized version of
goprotobuf) - github.com/DeDiS/protobuf (reflection based)
- github.com/google/flatbuffers
go get -u -t
go test -bench='.*' ./Shameless plug: I use pawk to format the table:
go test -bench='.*' ./ | pawk -F'\t' '"%-40s %10s %10s %s %s" % f'If performance, correctness and interoperability are the most important factors, gogoprotobuf is currently the best choice. It does require a pre-processing step (eg. via Go 1.4's "go generate" command).
But as always, make your own choice based on your requirements.
The data being serialized is the following structure with randomly generated values:
type A struct {
Name string
BirthDay time.Time
Phone string
Siblings int
Spouse bool
Money float64
}Results with Go 1.3 on an Intel i7-3930K:
benchmark iter time/iter bytes alloc allocs
--------- ---- --------- ----------- ------
BenchmarkVmihailencoMsgpackMarshal 1000000 2376 ns/op 352 B/op 5 allocs/op
BenchmarkVmihailencoMsgpackUnmarshal 1000000 2380 ns/op 347 B/op 13 allocs/op
BenchmarkJsonMarshal 500000 3501 ns/op 584 B/op 7 allocs/op
BenchmarkJsonUnmarshal 300000 5553 ns/op 447 B/op 8 allocs/op
BenchmarkBsonMarshal 500000 2347 ns/op 504 B/op 19 allocs/op
BenchmarkBsonUnmarshal 500000 2596 ns/op 296 B/op 22 allocs/op
BenchmarkVitessBsonMarshal 1000000 1735 ns/op 1168 B/op 4 allocs/op
BenchmarkVitessBsonUnmarshal 1000000 1036 ns/op 224 B/op 4 allocs/op
BenchmarkGobMarshal 200000 10662 ns/op 1688 B/op 35 allocs/op
BenchmarkGobUnmarshal 30000 49289 ns/op 17493 B/op 377 allocs/op
BenchmarkXdrMarshal 500000 2821 ns/op 520 B/op 24 allocs/op
BenchmarkXdrUnmarshal 1000000 2163 ns/op 271 B/op 12 allocs/op
BenchmarkUgorjiCodecMsgpackMarshal 500000 4096 ns/op 1905 B/op 10 allocs/op
BenchmarkUgorjiCodecMsgpackUnmarshal 500000 3770 ns/op 1840 B/op 14 allocs/op
BenchmarkUgorjiCodecBincMarshal 300000 4066 ns/op 1938 B/op 10 allocs/op
BenchmarkUgorjiCodecBincUnmarshal 500000 4004 ns/op 2000 B/op 17 allocs/op
BenchmarkSerealMarshal 300000 4857 ns/op 1360 B/op 26 allocs/op
BenchmarkSerealUnmarshal 500000 4287 ns/op 972 B/op 37 allocs/op
BenchmarkBinaryMarshal 1000000 2440 ns/op 408 B/op 19 allocs/op
BenchmarkBinaryUnmarshal 1000000 2339 ns/op 416 B/op 24 allocs/op
BenchmarkMsgpMarshal 3000000 447 ns/op 144 B/op 1 allocs/op
BenchmarkMsgpUnmarshal 3000000 584 ns/op 112 B/op 3 allocs/op
BenchmarkGoprotobufMarshal 2000000 971 ns/op 312 B/op 4 allocs/op
BenchmarkGoprotobufUnmarshal 1000000 1220 ns/op 432 B/op 9 allocs/op
BenchmarkGogoprotobufMarshal 10000000 229 ns/op 64 B/op 1 allocs/op
BenchmarkGogoprotobufUnmarshal 5000000 323 ns/op 112 B/op 3 allocs/op
BenchmarkFlatbuffersMarshal 3000000 523 ns/op 0 B/op 0 allocs/op
BenchmarkFlatBuffersUnmarshal 30000000 53.4 ns/op 0 B/op 0 allocs/op
BenchmarkProtobufMarshal 1000000 1452 ns/op 232 B/op 9 allocs/op
BenchmarkProtobufUnmarshal 1000000 1111 ns/op 192 B/op 10 allocs/op
Note: the gob results are not really representative of normal performance, as gob is designed for serializing streams or vectors of a single type, not individual values.
The benchmarks can also be run with validation enabled.
VALIDATE=1 go test -bench='.*' ./Unfortunately, several of the serializers exhibit issues:
- (minor) BSON drops sub-microsecond precision from
time.Time. - (minor) Vitess BSON drops sub-microsecond precision from
time.Time. - (minor) Ugorji Binc Codec drops the timezone name (eg. "EST" -> "-0500") from
time.Time.
BenchmarkVmihailencoMsgpackMarshal 1000000 1770 ns/op 408 B/op 8 allocs/op
BenchmarkVmihailencoMsgpackUnmarshal 500000 3073 ns/op 512 B/op 15 allocs/op
BenchmarkJsonMarshal 500000 3292 ns/op 584 B/op 7 allocs/op
BenchmarkJsonUnmarshal 200000 6077 ns/op 543 B/op 10 allocs/op
BenchmarkBsonMarshal 1000000 2302 ns/op 504 B/op 19 allocs/op
BenchmarkBsonUnmarshal --- FAIL: BenchmarkBsonUnmarshal
serialization_benchmarks_test.go:299: unmarshaled object differed:
&{36f9dfc8311e1efd 2014-12-20 08:28:56.697060419 +1100 AEDT bce6c01982 0 true 0.9078469433531421}
&{36f9dfc8311e1efd 2014-12-20 08:28:56.697 +1100 AEDT bce6c01982 0 true 0.9078469433531421}
BenchmarkVitessBsonMarshal 1000000 1644 ns/op 1168 B/op 4 allocs/op
BenchmarkVitessBsonUnmarshal --- FAIL: BenchmarkVitessBsonUnmarshal
serialization_benchmarks_test.go:299: unmarshaled object differed:
&{d22635208bfe823f 2014-12-20 08:28:58.370539555 +1100 AEDT 6c6e6d4ef8 2 true 0.2026063151373092}
&{d22635208bfe823f 2014-12-19 21:28:58.37 +0000 UTC 6c6e6d4ef8 2 true 0.2026063151373092}
BenchmarkGobMarshal 200000 9770 ns/op 1688 B/op 35 allocs/op
BenchmarkGobUnmarshal 30000 47695 ns/op 17590 B/op 379 allocs/op
BenchmarkXdrMarshal 500000 2559 ns/op 519 B/op 24 allocs/op
BenchmarkXdrUnmarshal 500000 3016 ns/op 352 B/op 14 allocs/op
BenchmarkUgorjiCodecMsgpackMarshal 500000 3662 ns/op 1464 B/op 24 allocs/op
BenchmarkUgorjiCodecMsgpackUnmarshal 300000 4533 ns/op 1232 B/op 29 allocs/op
BenchmarkUgorjiCodecBincMarshal 500000 3745 ns/op 1480 B/op 24 allocs/op
BenchmarkUgorjiCodecBincUnmarshal --- FAIL: BenchmarkUgorjiCodecBincUnmarshal
serialization_benchmarks_test.go:299: unmarshaled object differed:
&{94fd6da7835a0346 2014-12-20 08:29:10.50608751 +1100 AEDT 7749978493 0 true 0.09633537464986415}
&{94fd6da7835a0346 2014-12-20 08:29:10.50608751 +1100 +1100 7749978493 0 true 0.09633537464986415}
BenchmarkSerealMarshal 300000 4572 ns/op 1360 B/op 26 allocs/op
BenchmarkSerealUnmarshal 300000 4950 ns/op 1068 B/op 39 allocs/op
BenchmarkBinaryMarshal 1000000 2168 ns/op 408 B/op 19 allocs/op
BenchmarkBinaryUnmarshal 500000 3249 ns/op 512 B/op 26 allocs/op
BenchmarkGoprotobufMarshal 2000000 928 ns/op 312 B/op 4 allocs/op
BenchmarkGoprotobufUnmarshal 1000000 1201 ns/op 432 B/op 9 allocs/op
BenchmarkProtobufMarshal 1000000 1342 ns/op 232 B/op 9 allocs/op
BenchmarkProtobufUnmarshal 1000000 2128 ns/op 288 B/op 12 allocs/op
All other fields are correct however.