On 20 April 2011 17:42, Mark Thomas <ma...@apache.org> wrote: > On 20/04/2011 17:33, Filip Hanik - Dev Lists wrote: >> On 4/20/2011 10:07 AM, Mark Thomas wrote: >>> The new code is better than the old. You are welcome to try and improve >>> it further although I don't think there is much more scope for >>> improvement given the performance figures I am seeing. >> My suggestion here would be to: >> 1. When a JAR file has been scanned, then save the result of the scan >> to the work directory containing >> - name of jar >> - timestamp of the JAR >> - MD5 of jar >> - size of jar in bytes >> - result of scan >> >> This would serve a jar scanner cache. If I restart tomcat several times >> without any changes to the libraries, there is no need to incur the 12 >> second startup time. While 30s -> 12s is a dramatic improvement, you can >> bring down the penalty of a scan even more. > > As per my previous mail, that 12s is nearly all app init time, *not* JAR > scanning time. From the current performance figures I'm far from > convinced that a cache would add any value.
And the disk cache files still have to be opened and read ... > > Mark > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org > For additional commands, e-mail: dev-h...@tomcat.apache.org > > --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
