On 28/06/2016 15:54, jean-frederic clere wrote: > On 06/28/2016 04:05 PM, ma...@apache.org wrote: >> + Correct a potential performance problem identified by Nate Clark due >> to >> + Tomcat Native providing OpenSSL with thread identifiers poorly suited >> to >> + the hash function used by OpenSSL when selecting a bucket for the hash >> + that holds the per thread error data. > > I had issue with the comparing APR connector + openssl versus NIO + > openssl can't be that related > (http://www.slideshare.net/jfclere/tomcat-openssl slide 19) I first sent > small file then big ones... APR +openssl doesn't look to go quite where > I excepted it.
Give the concurrency and (assuming you weren't testing on Windows) those results do look like you might have hit this problem. I don't suppose you are in a position to do a quick test with tc-native trunk to see if it is any better now? Cheers, Mark --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
