java – 为什么我看到接受新频道的Netty服务器引导程序的性能下降？
内容导读
互联网集市收集整理的这篇技术教程文章主要介绍了java – 为什么我看到接受新频道的Netty服务器引导程序的性能下降？，小编现在分享给大家，供广大互联网技能从业者学习和参考。文章包含6677字，纯文字阅读大概需要10分钟。
内容图文
我还在使用Netty 3.10.我写了一个单元测试来检查Netty boss线程的性能.我在单元测试主线程中使用一个简单的Netty服务器引导程序,并在缓存的线程池中生成100个Java sync-IO客户端.我注意到性能下降,我觉得很奇怪.每个客户端打开一个套接字,写入数据并关闭,关闭后记录持续时间(ms).我的单元测试已附上.我的单元测试的典型输出是按给定的顺序：
> 43 x客户完成.持续时间：0
> 26 x客户完成.持续时间：16
> 16 x客户完成.持续时间：0
> 3 x客户完成.持续时间：517
> 11 x客户完成.持续时间：3003
> 1 x客户完成.持续时间：6036
因此,有1个客户端必须等待6秒才能获得开放的TCP / IP通道,11个客户端必须等待3秒钟.我还检查了花费/丢失的时间.它始终是客户端的新Socket(String,int).在服务器端,当管道工厂被触发时,时间已经消失.
我的单元的线程模型是否测试了这个或真正的Netty bootstrap / boss的原因？
import java.net.InetSocketAddress;
import java.net.Socket;
import java.nio.charset.Charset;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.atomic.AtomicInteger;
import org.jboss.netty.bootstrap.ServerBootstrap;
import org.jboss.netty.channel.Channel;
import org.jboss.netty.channel.ChannelFactory;
import org.jboss.netty.channel.ChannelHandlerContext;
import org.jboss.netty.channel.ChannelPipeline;
import org.jboss.netty.channel.ChannelPipelineFactory;
import org.jboss.netty.channel.ChannelStateEvent;
import org.jboss.netty.channel.Channels;
import org.jboss.netty.channel.ExceptionEvent;
import org.jboss.netty.channel.MessageEvent;
import org.jboss.netty.channel.SimpleChannelHandler;
import org.jboss.netty.channel.group.ChannelGroup;
import org.jboss.netty.channel.group.DefaultChannelGroup;
import org.jboss.netty.channel.socket.nio.NioServerSocketChannelFactory;
import org.junit.After;
import org.junit.Assert;
import org.junit.Test;
public class NettyServerBossTest {
private static final String SRV_HOST = "localhost";
private static final int SRV_PORT = 8080;
private static final byte[] MSG = "Hello world!".getBytes(Charset.forName("UTF-8"));
private static final int WAIT_MAX_MILLIS = 10 * 1000;
private final ChannelGroup channels = new DefaultChannelGroup();
private final int expected = 100;
private final AtomicInteger actual = new AtomicInteger();
private volatile boolean failed;
private ExecutorService clientThreads;
private Throwable cause;
private ServerBootstrap bootstrap;
@Test
public void test() {
createServer();
createClients();
awaitClients();
verifyFailure();
}
private void awaitClients() {
final long startMillis = System.currentTimeMillis();
final long maxMillis = startMillis + WAIT_MAX_MILLIS;
while ((this.actual.get() < this.expected) && !isFailed() && (System.currentTimeMillis() < maxMillis)) {
try {
Thread.sleep(250L);
} catch (InterruptedException e) {
e.printStackTrace();
}
}
System.out.println("Total duration: " + (System.currentTimeMillis() - startMillis));
Assert.assertEquals(this.expected, this.actual.get());
}
private void createClients() {
this.clientThreads = Executors.newCachedThreadPool();
for (int i = 0; i < this.expected; i++) {
this.clientThreads.execute(new PlainSocketClient());
}
}
private void closeChannels() {
try {
this.channels.close().await(10000);
} catch (InterruptedException e) {
e.printStackTrace();
}
}
private void createServer() {
final ExecutorService bosses = Executors.newCachedThreadPool();
final ExecutorService workers = Executors.newCachedThreadPool();
final ChannelFactory factory = new NioServerSocketChannelFactory(bosses, workers);
this.bootstrap = new ServerBootstrap(factory);
this.bootstrap.setPipelineFactory(new ChannelPipelineFactory() {
@Override
public ChannelPipeline getPipeline() {
return Channels.pipeline(new DiscardServerHandler());
}
});
this.bootstrap.setOption("child.tcpNoDelay", Boolean.TRUE);
this.bootstrap.setOption("child.keepAlive", Boolean.TRUE);
this.bootstrap.bind(new InetSocketAddress(SRV_HOST, SRV_PORT));
}
/**
* Fail unit test
*
* @param cause
* cause of failure
*/
public synchronized void setCause(Throwable cause) {
if (!this.failed && (cause == null)) {
this.failed = true;
this.cause = cause;
}
}
@After
public void after() {
closeChannels();
if (this.clientThreads != null) {
this.clientThreads.shutdownNow();
}
if (this.bootstrap != null) {
this.bootstrap.releaseExternalResources();
}
}
/**
* Check if unit test has failed
*
* @return <code>true</code> if failed, <code>false</code> if still OK
*/
public boolean isFailed() {
return this.failed;
}
/**
* Get cause of failure
*
* @return cause or <code>null</code>
*/
public synchronized Throwable getCause() {
return this.cause;
}
/**
* Make sure test has not failed with exception
*/
public void verifyFailure() {
if (this.failed) {
throw new IllegalStateException("test failed", getCause());
}
}
public abstract class TestRunnable implements Runnable {
@Override
public final void run() {
try {
execute();
} catch (Exception e) {
handleException(e);
}
}
protected abstract void handleException(Throwable e);
protected abstract void execute() throws Exception;
}
public abstract class AsyncThreadsTestRunnable extends TestRunnable {
@Override
protected final void handleException(Throwable e) {
setCause(e);
}
}
public class PlainSocketClient extends AsyncThreadsTestRunnable {
@Override
protected void execute() throws Exception {
final long startMillis = System.currentTimeMillis();
try (Socket sock = new Socket(SRV_HOST, SRV_PORT)) {
sock.getOutputStream().write(MSG);
}
NettyServerBossTest.this.actual.incrementAndGet();
System.out.println("Client done. Duration: " + (System.currentTimeMillis() - startMillis));
}
}
public class DiscardServerHandler extends SimpleChannelHandler {
@Override
public void channelConnected(ChannelHandlerContext ctx, ChannelStateEvent e) {
NettyServerBossTest.this.channels.add(e.getChannel());
}
@Override
public void messageReceived(ChannelHandlerContext ctx, MessageEvent e) {
// discard
}
@Override
public void exceptionCaught(ChannelHandlerContext ctx, ExceptionEvent e) {
e.getCause().printStackTrace();
Channel ch = e.getChannel();
ch.close();
}
}
}
解决方法:
我认为你记录的时间不是全部花在开放套接字上,它花费在线程切换上,因为当线程A打开套接字然后CPU可能切换到线程B,然后当套接字打开时,CPU可能不会切换到线程A一次但在执行了许多其他线程之后.我已经更改你的PlainSocketClient来添加一个synchronized来确保减少线程切换的影响：
public class PlainSocketClient extends AsyncThreadsTestRunnable {
private static final String LOCK = "LOCK";
@Override
protected void execute() throws Exception {
synchronized (LOCK) {
final long startMillis = System.currentTimeMillis();
try (Socket sock = new Socket(SRV_HOST, SRV_PORT)) {
sock.getOutputStream().write(MSG);
}
NettyServerBossTest.this.actual.incrementAndGet();
System.out.println("Client done. Duration: " + (System.currentTimeMillis() - startMillis));
}
}
}
然后他们几乎只输出0或1.您可以自己进行测试.
它只是证明了线程切换的耗时,并不意味着您需要在代码中添加synchronized.
内容总结
以上是互联网集市为您收集整理的java – 为什么我看到接受新频道的Netty服务器引导程序的性能下降？全部内容，希望文章能够帮你解决java – 为什么我看到接受新频道的Netty服务器引导程序的性能下降？所遇到的程序开发问题。 如果觉得互联网集市技术教程内容还不错，欢迎将互联网集市网站推荐给程序员好友。
内容备注
版权声明：本文内容由互联网用户自发贡献，该文观点与技术仅代表作者本人。本站仅提供信息存储空间服务，不拥有所有权，不承担相关法律责任。如发现本站有涉嫌侵权/违法违规的内容， 请发送邮件至 gblab@vip.qq.com 举报，一经查实，本站将立刻删除。
内容手机端
扫描二维码推送至手机访问。
来源：【匿名】