Gary, yes versions matter. For cp149's stuff, I don't know what version of Log4j2 they used, what version of Logback and what version of the disruptor...
We should probably upgrade to 3.2 for trunk, btw, thanks for the reminder! On Mon, Aug 26, 2013 at 11:20 PM, Gary Gregory <garydgreg...@gmail.com>wrote: > Also, does the version of the disruptor jar matter? I see that we are not > on the latest version 3.2.0 for trunk. > > Gary > > > On Mon, Aug 26, 2013 at 6:47 AM, Remko Popma <remko.po...@gmail.com>wrote: > >> I took a look. Still a bit rough, but people have started to integrate >> the disruptor into Logback. >> Unfortunately cp149 did not mention the software versions used, what OS >> they ran the performance test on, or any detail on the hardware they used >> (number of cores would be nice to know...), so it is hard to say anything >> about their performance results. >> I re-arranged the ranking by total throughput (threads x >> throughput/thread) below. >> >> Observations: >> 1. Log4j2 Async Appender does very well (beats Log4j2 Async Loggers and >> Logback jactor in all multi-threaded scenarios but one) >> 2. Logback Async disruptor roughly equivalent to Log4j2 Async Loggers and >> Async Appender (but hard to tell) >> 3. in the 2-thread scenario Logback Async disruptor is much better than >> Log4j2 Async Loggers (strange... Noise?) >> 4. Logback jactor (non-disruptor) appenders only do reasonably well in 1 >> thread scenarios, performance degrades in multi-thread scenarios >> >> My guess is this was run on Windows, my Windows performance results have >> also been noisy and much less consistent than Unix results. >> (Which reminds me, I should re-run the tests as we've made performance >> improvements and fixed memory leaks...) >> >> Ranking in total throughput (threads x throughput/thread): >> 1. Log4j2: Async Appender (4 threads): 10,632,480 ops/sec. >> 2. Logback: Async disruptor Appender (1 thread): 9,993,043 ops/sec. >> 3. Log4j2: Loggers all async (4 threads): 9,922,628 ops/sec. >> 4. Logback: Async disruptor Appender (4 threads): 9,204,316 ops/sec. >> 5. Logback: Async jactor2 Appender (1 thread): 9,001,575 ops/sec. >> 6. Logback: Async jactor Appender (1 thread): 8,482,989 ops/sec. >> 7. Log4j2: Loggers all async (1 thread): 8,394,794 ops/sec. >> 8. Logback: Async disruptor Appender (2 threads): 8,207,580 ops/sec. >> 9. Log4j2: Async Appender (2 threads): 7,658,818 ops/sec. >> 10. Log4j2: Async Appender (1 thread): 7,408,055 ops/sec. >> 11. Logback: Async jactor2 Appender (4 threads): 5,363,908 ops/sec. >> 12. Log4j2: Loggers all async (2 threads): 4,860,704 ops/sec. >> 13. Logback: Async jactor Appender (4 threads): 4,637,032 ops/sec. >> 14. Logback: Async jactor2 Appender (2 threads): 3,478,812 ops/sec. >> 15. Logback: Async jactor Appender (2 threads): 2,973,170 ops/sec. >> >> >> On Monday, August 26, 2013, Ralph Goers wrote: >> >>> Remko - I thought you might want to look at this - >>> https://github.com/cp149/jactor-logger >>> >>> Ralph >>> >> > > > -- > E-Mail: garydgreg...@gmail.com | ggreg...@apache.org > Java Persistence with Hibernate, Second > Edition<http://www.manning.com/bauer3/> > JUnit in Action, Second Edition <http://www.manning.com/tahchiev/> > Spring Batch in Action <http://www.manning.com/templier/> > Blog: http://garygregory.wordpress.com > Home: http://garygregory.com/ > Tweet! http://twitter.com/GaryGregory >
