我们在这一系列尝试给大家演示把Micro中的各种组件组合起来，测试它们之间各自配合的性能
以下测试目前在单机上执行
- Go-Version: go version go1.12.6 darwin/amd64
- Go-Micro: v1.8.3
- 操作系统：MacOSX 10.14.5 (18F203)
- 内存：32G
- 处理器：2.2 GHz Intel Core i7
我们先对比目前Micro中提供有的transport，它们分别是Http（默认）、grpc、memory、quic、tcp（插件）、nats（插件）、rabbitmq（插件）、utp（插件）
其中quic因为其go包经常变动，修改麻烦，暂不提供测试。未来Micro会改成基于更加稳定的kcp协议，二者都是基于UDP开发的协议。
| - | 平均 (ms) |
中位 (ms) |
最大 (ms) |
最小 (ms) |
P90 (ms) |
P99 (ms) |
TPS |
|---|---|---|---|---|---|---|---|
| HttpTransport（1） | 148.780 | 113.004 | 802.520 | 0.437 | 302.865 | 396 | 664 |
| TcpTransport | 2.047 | 1.485 | 67.188 | 0.109 | 3.646 | 23 | 45310 |
| GRPCTransport | 3.158 | 2.572 | 38.030 | 0.190 | 5.201 | 20 | 30102 |
| GRPC-TCPTransport | 3.064 | 2.562 | 36.174 | 0.214 | 4.871 | 19 | 30911 |
| UTPTransport | 6.757 | 6.583 | 41.935 | 0.136 | 8.402 | 24 | 14388 |
| NATsTransport（2） | 20.113 | 18.007 | 1003.489 | 1.697 | 28.260 | 596 | 4865 |
| NATsTransport（3） | 14.364 | 13.280 | 80.938 | 1.474 | 21.289 | 59 | 6773 |
| RabbitMQTransport（4） | 127.019 | 22.259 | 5005.101 | 1.866 | 32.702 | 5004 | 700 |
- 1 http模式与具体服务器对Http请求的出入限制配置有关，待优化配置再测试
- 2 nats运行在本机Docker中，分配内存为2GB，CPU核心数为6核
- 3 nats运行在本机Docker中，分配内存为4GB，CPU核心数为6核，当适当增加内存为双倍时，性能有40%左右提升，但与tpc，grpc仍有不小差距。
- 4 rabbit运行在本机Docker中，分配内存为4GB，CPU核心数为6核，性能极差，需要调优再测试
上面的表格中可见，TCP>grpc≈grpc-tcp>utp>nats>http>rabbitMQ
具体数值在不同宿主机中有差异，视具体情况而定，但是总体趋势是有参考意义的。
$ cd http-transport
$ go run server.go
# 切换窗口到同目录
$ go run client.go -c 100 -n 100000
2019/08/11 21:47:31 concurrency: 100
requests per client: 1000
2019/08/11 21:47:31 message size: 677 bytes
2019/08/11 21:50:01 took 150601 ms for 100000 requests
2019/08/11 21:50:01 sent requests : 100000
2019/08/11 21:50:01 received requests : 100000
2019/08/11 21:50:01 received requests_OK : 99966
2019/08/11 21:50:01 throughput (TPS) : 664
2019/08/11 21:50:01 concurrency mean median max min p90 p99 TPS
2019/08/11 21:50:01 100 148780257ns 113003500ns 802520000ns 437000ns 395687500ns 302865000ns 664
2019/08/11 21:50:01 100 148.780ms 113.004ms 802.520ms 0.437ms 302.865ms 396ms 664$ cd tcp-transport
$ go run server.go
# 切换窗口到同目录
$ go run client.go -c 100 -n 100000
2019/08/11 17:13:25 concurrency: 100
requests per client: 1000
2019/08/11 17:13:25 message size: 677 bytes
2019/08/11 17:13:27 took 2207 ms for 100000 requests
2019/08/11 17:13:27 sent requests : 100000
2019/08/11 17:13:27 received requests : 100000
2019/08/11 17:13:27 received requests_OK : 100000
2019/08/11 17:13:27 throughput (TPS) : 45310
2019/08/11 17:13:27 concurrency mean median max min p90 p99 TPS
2019/08/11 17:13:27 100 2047433ns 1485000ns 67188000ns 109000ns 23027000ns 3646000ns 45310
2019/08/11 17:13:27 100 2.047ms 1.485ms 67.188ms 0.109ms 3.646ms 23ms 45310$ cd grpc-transport
$ go run server.go
# 切换窗口到同目录
$ go run client.go -c 100 -n 100000
2019/08/11 22:16:23 concurrency: 100
requests per client: 1000
2019/08/11 22:16:23 message size: 677 bytes
2019/08/11 22:16:26 took 3322 ms for 100000 requests
2019/08/11 22:16:26 sent requests : 100000
2019/08/11 22:16:26 received requests : 100000
2019/08/11 22:16:26 received requests_OK : 100000
2019/08/11 22:16:26 throughput (TPS) : 30102
2019/08/11 22:16:26 concurrency mean median max min p90 p99 TPS
2019/08/11 22:16:26 100 3157656ns 2572000ns 38030000ns 190000ns 19897000ns 5201000ns 30102
2019/08/11 22:16:26 100 3.158ms 2.572ms 38.030ms 0.190ms 5.201ms 20ms 30102$ cd grpc-tcp-transport
$ go run server.go
# 切换窗口到同目录
$ go run client.go -c 100 -n 100000
2019/08/11 22:21:20 concurrency: 100
requests per client: 1000
2019/08/11 22:21:20 message size: 677 bytes
2019/08/12 22:21:20 sent requests : 100000
2019/08/12 22:21:20 received requests : 100000
2019/08/12 22:21:20 received requests_OK : 100000
2019/08/12 22:21:20 throughput (TPS) : 30911
2019/08/12 22:21:20 concurrency mean median max min p90 p99 TPS
2019/08/12 22:21:20 100 3064103ns 2562000ns 36174000ns 214000ns 18531500ns 4871000ns 30911
2019/08/12 22:21:20 100 3.064ms 2.562ms 36.174ms 0.214ms 4.871ms 19ms 30911$ cd utp-transport
$ go run server.go
# 切换窗口到同目录
$ go run client.go -c 100 -n 100000
2019/08/11 23:33:58 concurrency: 100
requests per client: 1000
2019/08/11 23:33:58 message size: 677 bytes
2019/08/11 23:34:05 took 6950 ms for 100000 requests
2019/08/11 23:34:05 sent requests : 100000
2019/08/11 23:34:05 received requests : 100000
2019/08/11 23:34:05 received requests_OK : 100000
2019/08/11 23:34:05 throughput (TPS) : 14388
2019/08/11 23:34:05 concurrency mean median max min p90 p99 TPS
2019/08/11 23:34:05 100 6757423ns 6583000ns 41935000ns 136000ns 24452000ns 8402000ns 14388
2019/08/11 23:34:05 100 6.757ms 6.583ms 41.935ms 0.136ms 8.402ms 24ms 14388$ cd nats-transport
$ go run server.go
# 切换窗口到同目录
$ go run client.go -c 100 -n 100000
2019/08/11 00:01:08 concurrency: 100
requests per client: 1000
2019/08/14 00:11:56 message size: 677 bytes
2019/08/14 00:12:17 took 20552 ms for 100000 requests
2019/08/14 00:12:17 sent requests : 100000
2019/08/14 00:12:17 received requests : 100000
2019/08/14 00:12:17 received requests_OK : 100000
2019/08/14 00:12:17 throughput (TPS) : 4865
2019/08/14 00:12:17 concurrency mean median max min p90 p99 TPS
2019/08/14 00:12:17 100 20112930ns 18007000ns 1003489000ns 1697000ns 595746000ns 28260000ns 4865
2019/08/14 00:12:17 100 20.113ms 18.007ms 1003.489ms 1.697ms 28.260ms 596ms 48652019/08/14 00:29:22 concurrency: 100
requests per client: 1000
2019/08/14 00:29:22 message size: 677 bytes
2019/08/14 00:42:20 sent requests : 100000
2019/08/14 00:42:20 received requests : 100000
2019/08/14 00:42:20 received requests_OK : 97908
2019/08/14 00:42:20 throughput (TPS) : 700
2019/08/14 00:42:20 concurrency mean median max min p90 p99 TPS
2019/08/14 00:42:20 100 127019375ns 22259000ns 5005101000ns 1866000ns 5003709000ns 32702000ns 700
2019/08/14 00:42:20 100 127.019ms 22.259ms 5005.101ms 1.866ms 32.702ms 5004ms 700