Apache ShenYu基准测试

什么是Apache ShenYu

Apache ShenYu是一个异步的，高性能的，跨语言的，响应式的Java API 网关。

Apache ShenYu 基准测试的环境

Mock服务器配置

• CPU: 4核8线程 Intel Cascade Lake @ 3.0GHz
• RAM: 16G

Apache ShenYu Bootstrap配置

• CPU: 4核8线程 Intel Cascade Lake @ 3.0GHz
• RAM: 16G

ShenYu 版本

• ShenYu Admin: 2.6.0
• ShenYu Bootstrap: 2.6.0

测试工具

• wrk-4.2.0

测试工具下载

• https://github.com/wg/wrk

测试工具安装

yum -y install git  make gcc

git clone https://github.com/wg/wrk.git wrk

cd wrk

make

cp wrk /usr/local/bin/

测试工具安装位置

测试工具占用资源很少，安装在网关节点服务器。

测试用例说明

说明

• 使用 Mock 服务模拟一个平均响应时长20ms，响应报文约2k的接口
• 每次测试时长 3 分钟
• JDK版本: OpenJdk-1.8.0
• HTTP 请求端分别使用 NettyClient 和 WebClient 进行测试
• 日志级别为WARN
• Apache ShenYu Bootstrap部署模式：单机部署
• Apache ShenYu Admin部署在mock服务器上。

JVM配置

-Xmx 4g 
-Xms 4g 
-Xmn 1g 
-Xss 512k 
-XX: +DisableExplicitGC 
-XX: LargePageSizeInBytes=128m

公共配置

matchCache:
  selector:
    selectorEnabled: false
    initialCapacity: 10000 # initial capacity in cache
    maximumSize: 10000 # max size in cache
  rule:
    initialCapacity: 10000 # initial capacity in cache
    maximumSize: 65536 # max size in cache
trie:
  enabled: true
  childrenSize: 10000
  pathVariableSize: 1000
  pathRuleCacheSize: 1000
  matchMode: antPathMatch

netty:
  http:
  # set to false, user can custom the netty tcp server config.
  webServerFactoryEnabled: true
  selectCount: 1
  workerCount: 8
  accessLog: false
  serverSocketChannel:
    soRcvBuf: 87380
    soBackLog: 128
    soReuseAddr: false
    connectTimeoutMillis: 10000
    writeBufferHighWaterMark: 65536
    writeBufferLowWaterMark: 32768
    writeSpinCount: 16
    autoRead: false
    allocType: "pooled"
    messageSizeEstimator: 8
    singleEventExecutorPerGroup: true
  socketChannel:
    soKeepAlive: false
    soReuseAddr: false
    soLinger: -1
    tcpNoDelay: true
    soRcvBuf: 87380
    soSndBuf: 16384
    ipTos: 0
    allowHalfClosure: false
    connectTimeoutMillis: 10000
    writeBufferHighWaterMark: 65536
    writeBufferLowWaterMark: 32768
    writeSpinCount: 16
    autoRead: false
    allocType: "pooled"
    messageSizeEstimator: 8
    singleEventExecutorPerGroup: true

WebClient配置

httpclient:
  strategy: webClient # netty
  connectTimeout: 45000 # 45000
  responseTimeout: 3000 # 3000
  readerIdleTime: 3000 # 3000
  writerIdleTime: 3000 # 3000
  allIdleTime: 3000 # 3000
  readTimeout: 3000 # 3000
  writeTimeout: 3000 # 3000
  wiretap: false # false
  keepAlive: false # false
  maxInMemorySize: 1 # 1
  pool:
    type: ELASTIC # ELASTIC
    name: proxy # proxy
    maxConnections: 16 # 16
    acquireTimeout: 45000 # 45000
    maxIdleTime: 3000 # 3000

NettyClient配置

httpclient:
  strategy: netty # netty
  connectTimeout: 45000 # 45000
  responseTimeout: 3000 # 3000
  readerIdleTime: 3000 # 3000
  writerIdleTime: 3000 # 3000
  allIdleTime: 3000 # 3000
  readTimeout: 3000 # 3000
  writeTimeout: 3000 # 3000
  wiretap: false # false
  keepAlive: false # false
  maxInMemorySize: 1 # 1
  pool:
    type: ELASTIC # ELASTIC
    name: proxy # proxy
    maxConnections: 16 # 16
    acquireTimeout: 45000 # 45000
    maxIdleTime: 3000 # 3000

启动ShenYu和Mock服务

启动ShenYu Admin

tar -zxvf apache-shenyu-admin-2.6.0.tar.gz
cd apache-shenyu-admin-2.6.0
cd bin
./start.sh

启动ShenYu Bootstrap

tar -zxvf apache-shenyu-bootstrap-2.6.0.tar.gz
cd apache-shenyu-bootstrap-2.6.0
cd bin
./start.sh

启动Mock服务

nohup java -jar shenyu-examples-http.jar -server -Xmx4g -Xms4g -Xmn1g -Xss256k &

如何测试

测试mock服务

./wrk -t8 -c600 -T10s -d180s --latency http://ip:8189/test/requstBody

测试ShenYu Bootstrap

./wrk -t8 -c600 -T10s -d180s --latency http://localhost:9195/http/test/requestBody
./wrk -t8 -c800 -T10s -d180s --latency http://localhost:9195/http/test/requestBody
./wrk -t8 -c1000 -T10s -d180s --latency http://localhost:9195/http/test/requestBody
./wrk -t8 -c1200 -T10s -d180s --latency http://localhost:9195/http/test/requestBody

测试结果

直接访问后端

QPS	50% latency (ms)	75% latency (ms)	90% latency (ms)	99% latency (ms)	平均响应时间(ms)	最大响应时间(ms)
28998.20	19.81	23.78	28.26	41.24	20.92	402.90

netty

currency	QPS	50% latency (ms)	75% latency (ms)	90% latency (ms)	99% latency (ms)	平均响应时间(ms)	最大响应时间(ms)
600 并发	20472.95	19.37	25.36	32.89	69.92	22.09	1043.33
800 并发	20703.55	23.57	31.32	40.11	77.28	26.11	576.47
1000 并发	20979.91	29.21	37.86	47.23	80.91	31.20	860.55
1200 并发	21129.88	32.45	42.40	52.68	96.10	35.06	1070

webClient

currency	QPS	50% latency (ms)	75% latency (ms)	90% latency (ms)	99% latency (ms)	平均响应时间(ms)	最大响应时间(ms)
600 并发	18640.47	15.77	24.77	38.26	80.31	20.32	852.06
800 并发	18723.44	18.12	28.69	44.96	95.3	23.52	765.26
1000 并发	18928.99	19.99	31.42	49.09	108.84	25.93	1040
1200 并发	18965.37	22.10	34.62	54.48	122.31	28.66	1075