测试方法
为了对Ignite做一个基本了解,做了一个性能测试,测试方法也比较简单主要是针对client模式,因为这种方法和使用redis的方式特别像。测试方法很简单主要是下面几点:
- 不作参数优化,默认配置进行测试
- 在一台linux服务器上部署Ignite服务端,然后自己的笔记本作客户端
- 按1,10,20,50,100,200线程进行测试
测试环境说明
服务器:
[09:36:56] ver. 1.7.0#20160801-sha1:383273e3[09:36:56] OS: Linux 2.6.32-279.el6.x86_64 amd64[09:36:56] VM information: Java(TM) SE Runtime Environment 1.7.0_07-b10 Oracle Corporation Java HotSpot(TM) 64-Bit Server VM 23.3-b01[09:36:56] Configured plugins:[09:36:56] ^-- None[09:36:56] [09:36:56] Security status [authentication=off, tls/ssl=off]CPU:4核
内存8GB
网卡100M
虚拟机客户机:
[13:05:32] ver. 1.7.0#20160801-sha1:383273e3[13:05:32] OS: Windows 7 6.1 amd64[13:05:32] VM information: Java(TM) SE Runtime Environment 1.8.0_40-b26 Oracle Corporation Java HotSpot(TM) 64-Bit Server VM 25.40-b25[13:05:32] Initial heap size is 128MB (should be no less than 512MB, use -Xms512m -Xmx512m).[13:05:34] Configured plugins:[13:05:34] ^-- None[13:05:34] [13:05:35] Security status [authentication=off, tls/ssl=off][13:05:51] Performance suggestions for grid(fix if possible)[13:05:51] To disable, set -DIGNITE_PERFORMANCE_SUGGESTIONS_DISABLED=true[13:05:51] ^-- Decrease number of backups (set "backups" to 0)CPU:4核,i5-4210u
内存8GB
笔记本win7 64位
网卡:100M测试代码
package org.j2server.j2cache.cache.iginte;import java.util.Arrays;import org.apache.ignite.Ignite;import org.apache.ignite.IgniteCache;import org.apache.ignite.Ignition;import org.apache.ignite.cache.CacheMode;import org.apache.ignite.configuration.CacheConfiguration;import org.apache.ignite.configuration.IgniteConfiguration;import org.apache.ignite.spi.discovery.tcp.TcpDiscoverySpi;import org.apache.ignite.spi.discovery.tcp.ipfinder.vm.TcpDiscoveryVmIpFinder;public class IgniteTest {//测试的数据行数private static final Integer test_rows = 50000;private static final Integer thread_cnt = 10;private static final String cacheName = "Ignite Cache";private static Ignite ignite;private static boolean client_mode = false;static {getIgnite();}public static void main(String[] args) {MultiThread();}private static Ignite getIgnite() {if (ignite == null) {TcpDiscoverySpi spi = new TcpDiscoverySpi();TcpDiscoveryVmIpFinder ipFinder = new TcpDiscoveryVmIpFinder();ipFinder.setAddresses(Arrays.asList("192.168.49.204"));spi.setIpFinder(ipFinder);CacheConfiguration cacheConfiguration = new CacheConfiguration<String, DataClass>();cacheConfiguration.setCacheMode(CacheMode.PARTITIONED);cacheConfiguration.setBackups(1);IgniteConfiguration cfg = new IgniteConfiguration();cfg.setClientMode(client_mode);cfg.setDiscoverySpi(spi);cfg.setCacheConfiguration(cacheConfiguration);ignite = Ignition.start(cfg);}System.out.println("是否客户端模式:" + client_mode);return ignite;}private static void MultiThread() {System.out.println("==================================================================");System.out.println("开始测试多线程写入[线程数:"+thread_cnt+"]");Long startTime = System.currentTimeMillis();Thread[] threads = new Thread[thread_cnt];Ignite ignite = getIgnite();IgniteCache<String, DataClass> cache = ignite.getOrCreateCache(cacheName);for (int i = 0; i < threads.length; i++) {threads[i] = new Thread(new TestThread(true, cache));}for (int i = 0; i< threads.length; i++) {threads[i].start();}for(Thread thread : threads){try {thread.join();} catch (InterruptedException e) { e.printStackTrace();}}Long endTime=System.currentTimeMillis(); //获取结束时间 float interval = endTime-startTime == 0 ? 1 : endTime-startTime;float tpms = (float)test_rows/interval;System.out.println("程序运行时间: "+ interval+"ms");System.out.println("每毫秒写入:"+tpms+"条。");System.out.println("每秒写入:"+tpms*1000+"条。"); System.out.println("==================================================================");System.out.println("开始测试多线程读取[线程数:"+thread_cnt+"]");startTime = System.currentTimeMillis();Thread[] readthreads = new Thread[thread_cnt];for (int i = 0; i < readthreads.length; i++) {readthreads[i] = new Thread(new TestThread(false, cache));}for (int i = 0; i< readthreads.length; i++) {readthreads[i].start();}for(Thread thread : readthreads){try {thread.join();} catch (InterruptedException e) { e.printStackTrace();}}endTime=System.currentTimeMillis(); //获取结束时间interval = endTime-startTime == 0 ? 1 : endTime-startTime;tpms = (float)test_rows/interval;System.out.println("程序运行时间: "+ interval+"ms");System.out.println("每毫秒读取:"+tpms+"条。");System.out.println("每秒读取:"+tpms*1000+"条。");}static class TestThread implements Runnable {private boolean readMode = true;private IgniteCache<String, DataClass> cache;public TestThread(boolean readMode, IgniteCache<String, DataClass> cache){this.readMode = readMode;this.cache = cache;}@Overridepublic void run() {for (int i = 0; i < test_rows/thread_cnt; i++) {if (this.readMode) {cache.get(Integer.toString(i));} else {DataClass dc = new DataClass();dc.setName(Integer.toString(i));dc.setValue(i);dc.setStrValue("asdfadsfasfda");cache.put(Integer.toString(i), dc);}}}}}import java.io.Serializable;public class DataClass implements Serializable{private String name;private long value;private String strValue;public String getName() {return name;}public void setName(String name) {this.name = name;}public long getValue() {return value;}public void setValue(long value) {this.value = value;}public String getStrValue() {return strValue;}public void setStrValue(String strValue) {this.strValue = strValue;}}测试数据
最终测试的结果还是有点意思,随着线程的增长读写性能大幅提升,但是到了200的时候就开始下降。下面是测试数据:
[12:53:40] Topology snapshot [ver=20, servers=1, clients=1, CPUs=8, heap=2.8GB]==================================================================开始测试多线程写入[线程数:1]程序运行时间: 49066.0ms每毫秒写入:1.0190356条。每秒写入:1019.0356条。==================================================================开始测试多线程读取[线程数:1]程序运行时间: 51739.0ms每毫秒读取:0.966389条。每秒读取:966.389条。[12:56:22] Topology snapshot [ver=22, servers=1, clients=1, CPUs=8, heap=2.8GB]==================================================================开始测试多线程写入[线程数:10]程序运行时间: 6215.0ms每毫秒写入:8.045053条。每秒写入:8045.0527条。==================================================================开始测试多线程读取[线程数:10]程序运行时间: 6526.0ms每毫秒读取:7.661661条。每秒读取:7661.661条。[12:57:04] Topology snapshot [ver=24, servers=1, clients=1, CPUs=8, heap=2.8GB]==================================================================开始测试多线程写入[线程数:20]程序运行时间: 4353.0ms每毫秒写入:11.486331条。每秒写入:11486.331条。==================================================================开始测试多线程读取[线程数:20]程序运行时间: 3768.0ms每毫秒读取:13.269639条。每秒读取:13269.639条。[12:57:34] Topology snapshot [ver=26, servers=1, clients=1, CPUs=8, heap=2.8GB]==================================================================开始测试多线程写入[线程数:50]程序运行时间: 2657.0ms每毫秒写入:18.818216条。每秒写入:18818.217条。==================================================================开始测试多线程读取[线程数:50]程序运行时间: 2138.0ms每毫秒读取:23.386343条。每秒读取:23386.344条。[12:58:00] Topology snapshot [ver=28, servers=1, clients=1, CPUs=8, heap=2.8GB]==================================================================开始测试多线程写入[线程数:100]程序运行时间: 2095.0ms每毫秒写入:23.866348条。每秒写入:23866.348条。==================================================================开始测试多线程读取[线程数:100]程序运行时间: 1764.0ms每毫秒读取:28.344671条。每秒读取:28344.672条。[12:59:19] Topology snapshot [ver=30, servers=1, clients=1, CPUs=8, heap=2.8GB]==================================================================开始测试多线程写入[线程数:200]程序运行时间: 2333.0ms每毫秒写入:21.431633条。每秒写入:21431.633条。==================================================================开始测试多线程读取[线程数:200]程序运行时间: 2049.0ms每毫秒读取:24.402147条。每秒读取:24402.146条。用图形看看比较直观不使用客户端模式
只不过我发现如果不使用client_mode,也就是都是server模式时写入性能还是很强的,但是读取有点搓。
[14:15:02] Topology snapshot [ver=22, servers=2, clients=0, CPUs=8, heap=2.8GB]是否客户端模式:false==================================================================开始测试多线程写入[线程数:1]是否客户端模式:false程序运行时间: 828.0ms每毫秒写入:60.386475条。每秒写入:60386.477条。==================================================================开始测试多线程读取[线程数:1]程序运行时间: 28819.0ms每毫秒读取:1.7349665条。每秒读取:1734.9666条。[14:08:55] Topology snapshot [ver=10, servers=2, clients=0, CPUs=8, heap=2.8GB]是否客户端模式:false==================================================================开始测试多线程写入[线程数:10]是否客户端模式:false程序运行时间: 813.0ms每毫秒写入:61.500614条。每秒写入:61500.613条。==================================================================开始测试多线程读取[线程数:10]程序运行时间: 5965.0ms每毫秒读取:8.38223条。每秒读取:8382.2295条。[14:09:48] Topology snapshot [ver=12, servers=2, clients=0, CPUs=8, heap=2.8GB]是否客户端模式:false==================================================================开始测试多线程写入[线程数:20]是否客户端模式:false程序运行时间: 812.0ms每毫秒写入:61.576355条。每秒写入:61576.355条。==================================================================开始测试多线程读取[线程数:20]程序运行时间: 5157.0ms每毫秒读取:9.6955595条。每秒读取:9695.56条。[14:10:25] Topology snapshot [ver=14, servers=2, clients=0, CPUs=8, heap=2.8GB]是否客户端模式:false==================================================================开始测试多线程写入[线程数:50]是否客户端模式:false程序运行时间: 686.0ms每毫秒写入:72.8863条。每秒写入:72886.3条。==================================================================开始测试多线程读取[线程数:50]程序运行时间: 4321.0ms每毫秒读取:11.571396条。每秒读取:11571.3955条。[14:11:01] Topology snapshot [ver=16, servers=2, clients=0, CPUs=8, heap=2.8GB]是否客户端模式:false==================================================================开始测试多线程写入[线程数:100]是否客户端模式:false程序运行时间: 830.0ms每毫秒写入:60.240963条。每秒写入:60240.965条。==================================================================开始测试多线程读取[线程数:100]程序运行时间: 3963.0ms每毫秒读取:12.616705条。每秒读取:12616.705条。[14:13:58] Topology snapshot [ver=20, servers=2, clients=0, CPUs=8, heap=2.8GB]是否客户端模式:false==================================================================开始测试多线程写入[线程数:200]是否客户端模式:false程序运行时间: 1014.0ms每毫秒写入:49.309666条。每秒写入:49309.664条。==================================================================开始测试多线程读取[线程数:200]程序运行时间: 3179.0ms每毫秒读取:15.728216条。每秒读取:15728.216条。用图形看看比较直观从这个数据可以看出来,在这种都是服务端的模式下,写入性能基本稳定,在达到200线程时出现衰减;而读取则基本是线性的,到100线程差不多也就到顶了。
与redis的对比
原本是想和redis作一个对比测试的,先是做了redis的测试。redis客户端用的jedis2.8.1,同时服务端用的是redis3.2.2,其他的环境和上面的一样。结果测试数据发现redis和ignite使用客户端模式时竟然很相近。所以我怀疑是因为我对redis不了解redis没作优化导致的?但是Ignite我也是直接启动的,一点优化也没作,还是说测试的代码写法不对呢?
下面是redis的测试代码import redis.clients.jedis.Jedis;public class redis {private static final String ip = "192.168.49.200";private static final String auth = "your pwd";private static final Integer port = 6379;//测试的数据行数private static final Integer test_rows = 50000;//线程数private static final Integer thread_cnt = 200;public static void main(String[] args) {MultiThread();}private static void MultiThread() {System.out.println("==================================================================");System.out.println("开始测试多线程写入[线程数:"+thread_cnt+"]");Long startTime = System.currentTimeMillis();Thread[] threads = new Thread[thread_cnt];for (int i = 0; i < threads.length; i++) {threads[i] = new Thread(new TestThread(true));}for (int i = 0; i< threads.length; i++) {threads[i].start();}for(Thread thread : threads){try {thread.join();} catch (InterruptedException e) { e.printStackTrace();}}Long endTime=System.currentTimeMillis(); //获取结束时间 float interval = endTime-startTime == 0 ? 1 : endTime-startTime;float tpms = (float)test_rows/interval;System.out.println("程序运行时间: "+ interval+"ms");System.out.println("每毫秒写入:"+tpms+"条。");System.out.println("每秒写入:"+tpms*1000+"条。"); System.out.println("==================================================================");System.out.println("开始测试多线程写入[线程数:"+thread_cnt+"]");startTime = System.currentTimeMillis();Thread[] readthreads = new Thread[thread_cnt];for (int i = 0; i < readthreads.length; i++) {readthreads[i] = new Thread(new TestThread(false));}for (int i = 0; i< readthreads.length; i++) {readthreads[i].start();}for(Thread thread : readthreads){try {thread.join();} catch (InterruptedException e) { e.printStackTrace();}}endTime=System.currentTimeMillis(); //获取结束时间interval = endTime-startTime == 0 ? 1 : endTime-startTime;tpms = (float)test_rows/interval;System.out.println("程序运行时间: "+ interval+"ms");System.out.println("每毫秒读取:"+tpms+"条。");System.out.println("每秒读取:"+tpms*1000+"条。");}static class TestThread implements Runnable {private boolean readMode = true;public TestThread(boolean readMode){this.readMode = readMode;}@Overridepublic void run() {Jedis j = new Jedis(ip,port);j.auth(auth);for (int i = 0; i < test_rows/thread_cnt; i++) {if (this.readMode) {j.get("foo"+i);} else {j.set("foo"+i, "bar"+i);}}j.disconnect();}}}对比结果视图结束
原本我想着redis估计得秒了ignite,毕竟redis是这么多系统正在使用的内存数据库。ignite本身含有这么多功能按理性能肯定是比不上才对,而且ignite组成集群后是需要进行数据分块存取和备份的,而测试环境中redis则是单实例情况,这让我没太想明白啊。。还望有高手指点。。看网上许多人测试的数据redis少点的4万+,据说可以到10万+。但我自己的测试环境差了点反正最多也没过3万,这到底会是什么原因呢?不管如何这是一次简单的测试与尝试,结果与预期有点偏差,继续学习深入了解吧。
本文永久更新链接地址