[logback-dev] [JIRA] (LOGBACK-1279) Async appenders do not scale well across multiple threads

QOS.CH (JIRA) noreply-jira at qos.ch
Mon Mar 13 13:15:00 CET 2017


Stephen Connolly created LOGBACK-1279:
-----------------------------------------

             Summary: Async appenders do not scale well across multiple threads
                 Key: LOGBACK-1279
                 URL: https://jira.qos.ch/browse/LOGBACK-1279
             Project: logback
          Issue Type: Bug
          Components: logback-classic
    Affects Versions: 1.2.1
            Reporter: Stephen Connolly
            Assignee: Logback dev list
         Attachments: image-2017-03-13-12-10-55-553.png

When running some benchmarking (specifically [https://github.com/stephenc/java-logging-benchmarks/blob/master/log4j-fmt/logback-async/src/main/java/me/justpitch/loggingbenchmarks/LoggingBenchmark.java] - at time of issue creation @ceki has access to repo, but repo is due to become public soon) I noticed that the Async appender does not scale well when the number of threads increases.

In fact the Async appender - which to my understanding is designed to help provide a more consistent latency under multiple threads - has dire performance when the number of threads increase, e.g. !image-2017-03-13-12-10-55-553.png!

The above results show that Log4J with an async appender has a profile very similar to that of both Logback and Log4J with a non-asynchronous appender. Logback's asynchronous appender performance gets progressively worse and approximates the performance of Java Utils Logging above 16 threads on the test system



--
This message was sent by Atlassian JIRA
(v7.3.1#73012)


More information about the logback-dev mailing list