The main takeaway for me is that the default should be to do buffered I/O. Unfortunately, I didn't see an option in Logback to enable buffering at all, much less make it the default. I'm doing that in Log4j2. Doing that cut my overall elapsed time in half. Ralph On Mar 20, 2011, at 9:31 AM, hostalp@post.cz wrote:
Yeah, also interesting stuff. And more just for the curiosity - I tried to wrap the FileWriter by BufferedWriter ( http://download.oracle.com/javase/6/docs/api/java/io/BufferedWriter.html ) in your test and to no surprise the result also became very close to BufferedOutputStream.
Following results are from 2nd runs within the JVM instance as first runs are (as usually) a little more "biased":
Java 6: FileOutputStream: 7186 BufferedOutputStream: 613 FileWriter: 713 BufferedWriter/FileWriter: 625
Java 1.5: FileOutputStream: 7765 BufferedOutputStream: 955 FileWriter: 1131 BufferedWriter/FileWriter: 918
Looks like the use of FileWriter itself is a decent option while adding buffering always helps at least a little. Also funny that BufferedWriter around FileWriter performs consistently better than BufferedOutputStream at Java 1.5 while at Java 6 it's exactly the opposite (but here the difference is minor). _______________________________________________ Logback-user mailing list Logback-user@qos.ch http://qos.ch/mailman/listinfo/logback-user