Benchmarks of Go serialization methods
This is a test suite for benchmarking various Go serialization methods.
Tested serialization methods
- encoding/gob
- encoding/json
- github.com/json-iterator/go
- 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
- labix.org/v2/mgo/bson
- github.com/tinylib/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
- github.com/hprose/hprose-go/io
- github.com/glycerine/go-capnproto
- zombiezen.com/go/capnproto2
- github.com/andyleap/gencode
- github.com/pascaldekloe/colfer
- github.com/linkedin/goavro
- github.com/ikkerens/ikeapack
- github.com/niubaoshu/gotiny
Running the benchmarks
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'Recommendation
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.
Data
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
2018-08-19 Results with Go 1.10 on a 2.5 GHz Intel Core i7 MacBook Pro (Retina, 15-inch, Mid 2015):
benchmark iter time/iter bytes/op allocs/op tt.time tt.bytes time/alloc
--------- ---- --------- -------- --------- ------- -------- ----------
BenchmarkGotinyMarshal-8 10000000 132 ns/op 0 B/op 0 allocs/op 1.32 s 0 KB 0.00 ns/alloc
BenchmarkGotinyUnmarshal-8 10000000 216 ns/op 112 B/op 3 allocs/op 2.16 s 112000 KB 72.00 ns/alloc
BenchmarkGotinyNoTimeMarshal-8 10000000 124 ns/op 0 B/op 0 allocs/op 1.24 s 0 KB 0.00 ns/alloc
BenchmarkGotinyNoTimeUnmarshal-8 10000000 203 ns/op 96 B/op 3 allocs/op 2.03 s 96000 KB 67.67 ns/alloc
BenchmarkMsgpMarshal-8 10000000 171 ns/op 128 B/op 1 allocs/op 1.71 s 128000 KB 171.00 ns/alloc
BenchmarkMsgpUnmarshal-8 5000000 326 ns/op 112 B/op 3 allocs/op 1.63 s 56000 KB 108.67 ns/alloc
BenchmarkVmihailencoMsgpackMarshal-8 1000000 1706 ns/op 368 B/op 6 allocs/op 1.71 s 36800 KB 284.33 ns/alloc
BenchmarkVmihailencoMsgpackUnmarshal-8 1000000 1863 ns/op 384 B/op 13 allocs/op 1.86 s 38400 KB 143.31 ns/alloc
BenchmarkJsonMarshal-8 1000000 2165 ns/op 528 B/op 5 allocs/op 2.17 s 52800 KB 433.00 ns/alloc
BenchmarkJsonUnmarshal-8 300000 4134 ns/op 495 B/op 8 allocs/op 1.24 s 14850 KB 516.75 ns/alloc
BenchmarkJsonIterMarshal-8 1000000 1259 ns/op 200 B/op 3 allocs/op 1.26 s 20000 KB 419.67 ns/alloc
BenchmarkJsonIterUnmarshal-8 1000000 1518 ns/op 248 B/op 7 allocs/op 1.52 s 24800 KB 216.86 ns/alloc
BenchmarkEasyJsonMarshal-8 1000000 1304 ns/op 784 B/op 5 allocs/op 1.30 s 78400 KB 260.80 ns/alloc
BenchmarkEasyJsonUnmarshal-8 1000000 1411 ns/op 160 B/op 4 allocs/op 1.41 s 16000 KB 352.75 ns/alloc
BenchmarkBsonMarshal-8 1000000 1373 ns/op 392 B/op 10 allocs/op 1.37 s 39200 KB 137.30 ns/alloc
BenchmarkBsonUnmarshal-8 1000000 1838 ns/op 244 B/op 19 allocs/op 1.84 s 24400 KB 96.74 ns/alloc
BenchmarkGobMarshal-8 2000000 942 ns/op 48 B/op 2 allocs/op 1.88 s 9600 KB 471.00 ns/alloc
BenchmarkGobUnmarshal-8 2000000 945 ns/op 112 B/op 3 allocs/op 1.89 s 22400 KB 315.00 ns/alloc
BenchmarkXdrMarshal-8 1000000 1660 ns/op 456 B/op 21 allocs/op 1.66 s 45600 KB 79.05 ns/alloc
BenchmarkXdrUnmarshal-8 1000000 1451 ns/op 240 B/op 11 allocs/op 1.45 s 24000 KB 131.91 ns/alloc
BenchmarkUgorjiCodecMsgpackMarshal-8 1000000 1072 ns/op 545 B/op 6 allocs/op 1.07 s 54500 KB 178.67 ns/alloc
BenchmarkUgorjiCodecMsgpackUnmarshal-8 1000000 1276 ns/op 465 B/op 6 allocs/op 1.28 s 46500 KB 212.67 ns/alloc
BenchmarkUgorjiCodecBincMarshal-8 1000000 1154 ns/op 577 B/op 7 allocs/op 1.15 s 57700 KB 164.86 ns/alloc
BenchmarkUgorjiCodecBincUnmarshal-8 1000000 1449 ns/op 689 B/op 9 allocs/op 1.45 s 68900 KB 161.00 ns/alloc
BenchmarkSerealMarshal-8 500000 2544 ns/op 912 B/op 21 allocs/op 1.27 s 45600 KB 121.14 ns/alloc
BenchmarkSerealUnmarshal-8 500000 2776 ns/op 1008 B/op 34 allocs/op 1.39 s 50400 KB 81.65 ns/alloc
BenchmarkBinaryMarshal-8 1000000 1403 ns/op 334 B/op 20 allocs/op 1.40 s 33400 KB 70.15 ns/alloc
BenchmarkBinaryUnmarshal-8 1000000 1507 ns/op 335 B/op 22 allocs/op 1.51 s 33500 KB 68.50 ns/alloc
BenchmarkFlatBuffersMarshal-8 5000000 319 ns/op 0 B/op 0 allocs/op 1.59 s 0 KB 0.00 ns/alloc
BenchmarkFlatBuffersUnmarshal-8 5000000 248 ns/op 112 B/op 3 allocs/op 1.24 s 56000 KB 82.67 ns/alloc
BenchmarkCapNProtoMarshal-8 3000000 475 ns/op 56 B/op 2 allocs/op 1.43 s 16800 KB 237.50 ns/alloc
BenchmarkCapNProtoUnmarshal-8 3000000 413 ns/op 200 B/op 6 allocs/op 1.24 s 60000 KB 68.83 ns/alloc
BenchmarkCapNProto2Marshal-8 2000000 696 ns/op 244 B/op 3 allocs/op 1.39 s 48800 KB 232.00 ns/alloc
BenchmarkCapNProto2Unmarshal-8 2000000 1016 ns/op 320 B/op 6 allocs/op 2.03 s 64000 KB 169.33 ns/alloc
BenchmarkHproseMarshal-8 2000000 897 ns/op 473 B/op 8 allocs/op 1.79 s 94600 KB 112.12 ns/alloc
BenchmarkHproseUnmarshal-8 1000000 1083 ns/op 320 B/op 10 allocs/op 1.08 s 32000 KB 108.30 ns/alloc
BenchmarkHprose2Marshal-8 2000000 664 ns/op 0 B/op 0 allocs/op 1.33 s 0 KB 0.00 ns/alloc
BenchmarkHprose2Unmarshal-8 3000000 536 ns/op 144 B/op 4 allocs/op 1.61 s 43200 KB 134.00 ns/alloc
BenchmarkProtobufMarshal-8 2000000 954 ns/op 200 B/op 7 allocs/op 1.91 s 40000 KB 136.29 ns/alloc
BenchmarkProtobufUnmarshal-8 2000000 835 ns/op 192 B/op 10 allocs/op 1.67 s 38400 KB 83.50 ns/alloc
BenchmarkGoprotobufMarshal-8 3000000 411 ns/op 96 B/op 2 allocs/op 1.23 s 28800 KB 205.50 ns/alloc
BenchmarkGoprotobufUnmarshal-8 2000000 624 ns/op 200 B/op 10 allocs/op 1.25 s 40000 KB 62.40 ns/alloc
BenchmarkGogoprotobufMarshal-8 10000000 153 ns/op 64 B/op 1 allocs/op 1.53 s 64000 KB 153.00 ns/alloc
BenchmarkGogoprotobufUnmarshal-8 10000000 225 ns/op 96 B/op 3 allocs/op 2.25 s 96000 KB 75.00 ns/alloc
BenchmarkColferMarshal-8 10000000 129 ns/op 64 B/op 1 allocs/op 1.29 s 64000 KB 129.00 ns/alloc
BenchmarkColferUnmarshal-8 10000000 178 ns/op 112 B/op 3 allocs/op 1.78 s 112000 KB 59.33 ns/alloc
BenchmarkGencodeMarshal-8 10000000 168 ns/op 80 B/op 2 allocs/op 1.68 s 80000 KB 84.00 ns/alloc
BenchmarkGencodeUnmarshal-8 10000000 181 ns/op 112 B/op 3 allocs/op 1.81 s 112000 KB 60.33 ns/alloc
BenchmarkGencodeUnsafeMarshal-8 20000000 105 ns/op 48 B/op 1 allocs/op 2.10 s 96000 KB 105.00 ns/alloc
BenchmarkGencodeUnsafeUnmarshal-8 10000000 140 ns/op 96 B/op 3 allocs/op 1.40 s 96000 KB 46.67 ns/alloc
BenchmarkXDR2Marshal-8 10000000 159 ns/op 64 B/op 1 allocs/op 1.59 s 64000 KB 159.00 ns/alloc
BenchmarkXDR2Unmarshal-8 10000000 140 ns/op 32 B/op 2 allocs/op 1.40 s 32000 KB 70.00 ns/alloc
BenchmarkGoAvroMarshal-8 500000 2419 ns/op 1030 B/op 31 allocs/op 1.21 s 51500 KB 78.03 ns/alloc
BenchmarkGoAvroUnmarshal-8 200000 5810 ns/op 3436 B/op 87 allocs/op 1.16 s 68720 KB 66.78 ns/alloc
BenchmarkGoAvro2TextMarshal-8 500000 2740 ns/op 1326 B/op 20 allocs/op 1.37 s 66300 KB 137.00 ns/alloc
BenchmarkGoAvro2TextUnmarshal-8 500000 2650 ns/op 807 B/op 34 allocs/op 1.32 s 40350 KB 77.94 ns/alloc
BenchmarkGoAvro2BinaryMarshal-8 2000000 859 ns/op 510 B/op 11 allocs/op 1.72 s 102000 KB 78.09 ns/alloc
BenchmarkGoAvro2BinaryUnmarshal-8 2000000 914 ns/op 576 B/op 13 allocs/op 1.83 s 115200 KB 70.31 ns/alloc
BenchmarkIkeaMarshal-8 2000000 614 ns/op 72 B/op 8 allocs/op 1.23 s 14400 KB 76.75 ns/alloc
BenchmarkIkeaUnmarshal-8 2000000 782 ns/op 160 B/op 11 allocs/op 1.56 s 32000 KB 71.09 ns/alloc
---
totals:
BenchmarkGencodeUnsafe-8 30000000 245 ns/op 144 B/op 4 allocs/op 7.35 s 432000 KB 61.25 ns/alloc
BenchmarkXDR2-8 20000000 299 ns/op 96 B/op 3 allocs/op 5.98 s 192000 KB 99.67 ns/alloc
BenchmarkColfer-8 20000000 307 ns/op 176 B/op 4 allocs/op 6.14 s 352000 KB 76.75 ns/alloc
BenchmarkGotinyNoTime-8 20000000 327 ns/op 96 B/op 3 allocs/op 6.54 s 192000 KB 109.00 ns/alloc
BenchmarkGotiny-8 20000000 348 ns/op 112 B/op 3 allocs/op 6.96 s 224000 KB 116.00 ns/alloc
BenchmarkGencode-8 20000000 349 ns/op 192 B/op 5 allocs/op 6.98 s 384000 KB 69.80 ns/alloc
BenchmarkGogoprotobuf-8 20000000 378 ns/op 160 B/op 4 allocs/op 7.56 s 320000 KB 94.50 ns/alloc
BenchmarkMsgp-8 15000000 497 ns/op 240 B/op 4 allocs/op 7.46 s 360000 KB 124.25 ns/alloc
BenchmarkFlatBuffers-8 10000000 567 ns/op 112 B/op 3 allocs/op 5.67 s 112000 KB 189.00 ns/alloc
BenchmarkCapNProto-8 6000000 888 ns/op 256 B/op 8 allocs/op 5.33 s 153600 KB 111.00 ns/alloc
BenchmarkGoprotobuf-8 5000000 1035 ns/op 296 B/op 12 allocs/op 5.17 s 148000 KB 86.25 ns/alloc
BenchmarkHprose2-8 5000000 1200 ns/op 144 B/op 4 allocs/op 6.00 s 72000 KB 300.00 ns/alloc
BenchmarkIkea-8 4000000 1396 ns/op 232 B/op 19 allocs/op 5.58 s 92800 KB 73.47 ns/alloc
BenchmarkCapNProto2-8 4000000 1712 ns/op 564 B/op 9 allocs/op 6.85 s 225600 KB 190.22 ns/alloc
BenchmarkGoAvro2Binary-8 4000000 1773 ns/op 1086 B/op 24 allocs/op 7.09 s 434400 KB 73.88 ns/alloc
BenchmarkProtobuf-8 4000000 1789 ns/op 392 B/op 17 allocs/op 7.16 s 156800 KB 105.24 ns/alloc
BenchmarkGob-8 4000000 1887 ns/op 160 B/op 5 allocs/op 7.55 s 64000 KB 377.40 ns/alloc
BenchmarkHprose-8 3000000 1980 ns/op 793 B/op 18 allocs/op 5.94 s 237900 KB 110.00 ns/alloc
BenchmarkUgorjiCodecMsgpack-8 2000000 2348 ns/op 1010 B/op 12 allocs/op 4.70 s 202000 KB 195.67 ns/alloc
BenchmarkUgorjiCodecBinc-8 2000000 2603 ns/op 1266 B/op 16 allocs/op 5.21 s 253200 KB 162.69 ns/alloc
BenchmarkEasyJson-8 2000000 2715 ns/op 944 B/op 9 allocs/op 5.43 s 188800 KB 301.67 ns/alloc
BenchmarkJsonIter-8 2000000 2777 ns/op 448 B/op 10 allocs/op 5.55 s 89600 KB 277.70 ns/alloc
BenchmarkBinary-8 2000000 2910 ns/op 669 B/op 42 allocs/op 5.82 s 133800 KB 69.29 ns/alloc
BenchmarkXdr-8 2000000 3111 ns/op 696 B/op 32 allocs/op 6.22 s 139200 KB 97.22 ns/alloc
BenchmarkBson-8 2000000 3211 ns/op 636 B/op 29 allocs/op 6.42 s 127200 KB 110.72 ns/alloc
BenchmarkVmihailencoMsgpack-8 2000000 3569 ns/op 752 B/op 19 allocs/op 7.14 s 150400 KB 187.84 ns/alloc
BenchmarkSereal-8 1000000 5320 ns/op 1920 B/op 55 allocs/op 5.32 s 192000 KB 96.73 ns/alloc
BenchmarkGoAvro2Text-8 1000000 5390 ns/op 2133 B/op 54 allocs/op 5.39 s 213300 KB 99.81 ns/alloc
BenchmarkJson-8 1300000 6299 ns/op 1023 B/op 13 allocs/op 8.19 s 132990 KB 484.54 ns/alloc
BenchmarkGoAvro-8 700000 8229 ns/op 4466 B/op 118 allocs/op 5.76 s 312620 KB 69.74 ns/alloc
Issues
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) Ugorji Binc Codec drops the timezone name (eg. "EST" -> "-0500") from
time.Time.
--- FAIL: BenchmarkBsonUnmarshal-8
serialization_benchmarks_test.go:115: unmarshaled object differed:
&{20b999e3621bd773 2016-01-19 14:05:02.469416459 -0800 PST f017c8e9de 4 true 0.20887343719329818}
&{20b999e3621bd773 2016-01-19 14:05:02.469 -0800 PST f017c8e9de 4 true 0.20887343719329818}
--- FAIL: BenchmarkUgorjiCodecBincUnmarshal-8
serialization_benchmarks_test.go:115: unmarshaled object differed:
&{20a1757ced6b488e 2016-01-19 14:05:15.69474534 -0800 PST 71f3bf4233 0 false 0.8712180830484527}
&{20a1757ced6b488e 2016-01-19 14:05:15.69474534 -0800 -0800 71f3bf4233 0 false 0.8712180830484527}
All other fields are correct however.
Additionally, while not a correctness issue, FlatBuffers, ProtoBuffers, Cap'N'Proto and ikeapack do not support time types directly. In the benchmarks an int64 value is used to hold a UnixNano timestamp.