环境
jdk:1.7.0_79
cpu:i5-4570@3.20GHz 4核
eclipse:3.7
操作系统:win7
准备
log4j:1.7.21
org.slf4j
slf4j-log4j12
1.7.21
log4j.xml
logback:1.1.7
ch.qos.logback
logback-classic
1.1.7
logback.xml
%d{HH:mm:ss.SSS} [%thread] %-5level %logger{36} - %msg%n
testFile.log
true
[%t] - %m%n
0
additivity="false">
log4j2:2.6.2
org.apache.logging.log4j
log4j-core
2.6.2
org.apache.logging.log4j
log4j-slf4j-impl
2.6.2
com.lmax
disruptor
3.3.4
log4j2.xml
filePattern="${sys:user.home}/logs/hpaasvc/$${date:yyyy-MM}/info-%d{yyyy-MM-dd}-%i.log">
onMismatch="NEUTRAL" />
immediateFlush="false" append="true">
[%t] - %m%n
测试
准备50条线程同时记录1000000条数据,然后统计时间,详细代码如下:
import java.util.concurrent.CountDownLatch;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
public class App {
private static Logger log = LoggerFactory.getLogger(App.class);
public static void main(String[] args) throws InterruptedException {
int messageSize = 1000000;
int threadSize = 50;
final int everySize = messageSize / threadSize;
final CountDownLatch cdl = new CountDownLatch(threadSize);
long startTime = System.currentTimeMillis();
for (int ts = 0; ts < threadSize; ts++) {
new Thread(new Runnable() {
@Override
public void run() {
for (int es = 0; es < everySize; es++) {
log.info("======info");
}
cdl.countDown();
}
}).start();
}
cdl.await();
long endTime = System.currentTimeMillis();
System.out.println("log4j1:messageSize = " + messageSize
+ ",threadSize = " + threadSize + ",costTime = "
+ (endTime - startTime) + "ms");
}
}
log4j1和logback的同步和异步分别修改为对应的appender就行了
log4j2的异步方式提供了2中模式:
1.全局开启
设置Log4jContextSelector系统属性为:
org.apache.logging.log4j.core.async.AsyncLoggerContextSelector
System.setProperty("Log4jContextSelector", "org.apache.logging.log4j.core.async.AsyncLoggerContextSelector");
2.混合同步异步模式
不需要设置Log4jContextSelector,但是需要使用AsyncLogger标签
更多详细参考官方文档:http://logging.apache.org/log4j/2.x/manual/async.html#AllAsync
结果
分别测试完以后统计成表格如下:
log4j2的异步模式表现了绝对的性能优势,优势主要得益于Disruptor框架的使用
LMAX Disruptor technology. Asynchronous Loggers internally use the Disruptor, a lock-free inter-thread communication library, instead of queues, resulting in higher throughput and lower latency.
一个无锁的线程间通信库代替了原来的队列
更多Disruptor :
http://developer.51cto.com/art/201306/399370.htm