RPC Benchmark 测试: 将开源RPC框架聚在一起是一种怎样的体验？

本文转载自 RPC Benchmark 作者 鲁小憨 的简书博文 

https://www.jianshu.com/p/caf51f5cfbaa

测试说明

• 仅限于 Java
• 客户端使用 JMH 进行压测, 32 线程, 3 轮预热 3 轮测试, 每轮 10s
• 每次运行前都会执行 killall java, 但没有在每轮测试时重启操作系统
• 所有类库版本在发布时都是最新的, 除非存在 bug
• 所有框架都尽量参考该项目自带的 Benchmark 实现
• 将会一直持续, 不定期发布测试结果

测试用例

1. boolean existUser(String email), 判断某个 email 是否存在
2. boolean createUser(User user), 添加一个 User
3. User getUser(long id), 根据 id 获取一个用户
4. Page<User> listUser(int pageNo), 获取用户列表

运行结果

生成时间: 2018-05-12 21:15:11
硬件环境: 阿里云 ecs.hfc5.xlarge Intel Xeon Gold 6149, 4CPU 8GB RAM 两台
软件环境: Ubuntu x64 16.04.5, Java HotSpot(TM) 64-Bit Server VM 10+46
启动参数: java -server -Xmx1g -Xms1g -XX:MaxDirectMemorySize=1g -XX:+UseG1GC

existUser

framework	thrpt (ops/ms)	avgt (ms)	p90 (ms)	p99 (ms)	p999 (ms)
turbo-rpc	152.012	0.21	0.318	0.447	0.669
servicecomb	129.413	0.246	0.294	0.358	1.279
jupiter	128.015	0.249	0.36	0.543	3.826
thrift	117.827	0.273	0.52	0.97	1.454
netty	102.939	0.304	0.335	0.386	0.815
turbo-rest	95.774	0.341	0.845	1.505	4.047
undertow	92.956	0.342	0.88	1.552	3.532
undertow-async	82.539	0.38	0.773	1.604	5.153
armeria-http	75.795	0.411	0.731	1.305	6.054
dubbo	74.731	0.43	0.532	0.7	3.361
dubbo-kryo	74.728	0.43	0.532	0.713	3.758
motan	61.324	0.516	0.813	2.013	11.862
rapidoid	58.35	0.554	1.634	3.26	10.568
hprose	45.945	0.735	0.414	2.39	40.239
springboot-undertow	34.374	0.938	1.135	8.764	21.561
springwebflux	32.075	1.008	1.571	1.995	8.249
grpc	31.134	1.065	1.294	1.509	6.971
springboot	26.885	1.185	1.599	9.945	23.298

createUser

framework	thrpt (ops/ms)	avgt (ms)	p90 (ms)	p99 (ms)	p999 (ms)
servicecomb	124.015	0.255	0.306	0.382	1.436
jupiter	114.996	0.275	0.412	0.633	4.002
turbo-rpc	114.837	0.282	0.376	0.543	0.823
thrift	110.952	0.294	0.563	1.041	1.606
undertow	93.593	0.346	0.823	1.479	3.903
turbo-rest	88.512	0.353	0.894	1.618	5.087
netty	85.774	0.388	0.428	0.511	0.791
undertow-async	79.475	0.409	0.8	1.528	5.235
armeria-http	65.109	0.498	0.688	1.98	7.922
motan	56.708	0.561	0.893	1.892	11.289
dubbo-kryo	54.754	0.583	0.732	0.938	5.169
hprose	37.22	0.933	0.484	38.601	43.451
dubbo	31.811	0.988	1.284	1.642	7.471
springwebflux	30.212	1.084	1.694	2.13	6.824
grpc	29.992	1.029	1.321	1.544	4.5
springboot-undertow	28.142	1.139	2.003	6.398	13.976
springboot	26.411	1.196	1.647	10.994	26.771
rapidoid	22.434	1.438	0.913	26.345	46.137

getUser

framework	thrpt (ops/ms)	avgt (ms)	p90 (ms)	p99 (ms)	p999 (ms)
turbo-rpc	133.121	0.241	0.364	0.516	0.743
servicecomb	122.801	0.262	0.313	0.391	2.651
jupiter	111.682	0.29	0.45	0.689	4.301
thrift	110.368	0.289	0.553	1.02	1.573
undertow	87.382	0.343	0.981	1.765	3.945
turbo-rest	81.373	0.392	0.696	1.397	4.383
undertow-async	81.162	0.405	0.773	1.534	3.777
netty	73.999	0.421	0.442	0.472	0.886
armeria-http	70.776	0.448	0.753	1.389	4.366
motan	60.062	0.542	0.87	1.964	11.813
dubbo-kryo	55.375	0.564	0.722	0.951	5.317
rapidoid	51.64	0.625	2.019	4.076	11.72
hprose	48.659	0.649	0.701	1.356	40.042
dubbo	32.924	0.974	1.245	1.692	7.406
grpc	30.941	1.042	1.233	1.462	3.453
springboot-undertow	30.694	1.063	1.27	9.241	21.463
springwebflux	26.173	1.232	1.915	2.359	6.955
springboot	24.969	1.297	1.743	10.879	27.329

listUser

framework	thrpt (ops/ms)	avgt (ms)	p90 (ms)	p99 (ms)	p999 (ms)
jupiter	46.89	0.684	1.057	1.8	9.273
turbo-rpc	45.852	0.713	1.745	2.126	7.274
servicecomb	45.218	0.707	0.977	1.718	6.94
thrift	34.813	0.91	1.78	3.727	9.208
grpc	34.112	0.916	1.2	1.491	6.151
motan	32.584	0.982	1.632	2.998	10.928
undertow	32.094	1.01	1.352	2.171	9.29
undertow-async	32.094	0.994	1.298	2.494	10.943
armeria-http	31.701	1.018	1.579	2.9	9.945
netty	26.653	1.18	1.335	1.743	4.538
rapidoid	23.563	1.368	1.128	20.873	38.076
turbo-rest	21.274	1.516	1.823	3.166	9.929
dubbo-kryo	20.664	1.553	2.087	2.982	8.716
hprose	20.017	1.602	1.597	2.044	50.397
springwebflux	19.617	1.631	2.56	3.084	9.077
springboot-undertow	18.478	1.745	2.908	11.944	21.758
springboot	16.825	1.909	2.585	13.23	34.275
dubbo	4.683	6.922	9.224	12.534	19.741

更新说明

1. 硬件变化, 升级为 ecs.hfc5.xlarge Intel Xeon Gold 6149
2. 添加 3 个新的测评项目：armeria servicecomb springboot-undertow
3. 其他: 常规版本升级, 都升级到了最新版本 (dubbo 2.6.1 存在 bug，继续使用 2.6.0)

致谢

特别感谢下列人员对本项目的大力支持

1. turbo: hank-whu
2. jupiter: fengjiachun
3. hprose: andot
4. servicecomb: imlidian

免责声明

• 能力所限错误在所难免, 本测试用例及测试结果仅供参考
• 如果你认为 xxx 框架的代码或配置存在问题，那么欢迎发起 Pull Request
• 利益相关: 本测试用例作者同时为 turbo undertow-async 的作者

快速链接

往期评测:
RPC Benchmark Round 2
RPC Benchmark Round 1

测试说明:
rpc-benchmark 测试说明
rpc-benchmark 项目代码