If you need to serve static files for a high volume website, you're better off paying a specialty provider for it. Back when I worked at verizon, we used Akamai for static files like images etc.
serving up a ton of large static files quickly swamps your bandwidth, so the question isn't whether that 4% matters. It's really about how much bandwidth you're chewing up with static data. Unless the servers are hosted at a Tier 1 provider with OC12 bandwidth or higher, it really isn't going to make any difference from my experience. peter On Mon, May 18, 2009 at 4:34 PM, Robin Wilson <rwil...@kingsisle.com> wrote: > I don't know if I'd call a 4% difference a "dead heat"... I guess that would > depend on how many of those files you are serving a day... If I had 25 > servers all working full-throttle all day, 4% would be enough to require 1 > more server. If my peak load exceeds the necessary threshold, 4% could mean I > get end-users seeing errors periodically through the day - even with much > smaller configurations. > > Just something to think about... > > -- > Robin D. Wilson > Director of Web Development > KingsIsle Entertainment, Inc. > WORK: 512-623-5913 > CELL: 512-426-3929 > www.KingsIsle.com > > > > -----Original Message----- > From: Christopher Schultz [mailto:ch...@christopherschultz.net] > Sent: Monday, May 18, 2009 3:25 PM > To: Tomcat Users List > Subject: Re: Apache httpd vs Tomcat static content performance > [Revised/Updated] > > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Robin, > > On 5/18/2009 4:11 PM, Robin Wilson wrote: >> Thanks! This information isn't useless... Of course, more detailed >> results, after a longer test run would be more conclusive. > > Yup, that's the plan. Tonight, I'll be running with an 8 minute test to > give me 12 solid hours of testing. /Those/ should be more definitive > results. I've also rigged my test to prime the server by hitting each > file a single time, then waiting a few seconds, then starting the real test. > >> This appears to show that Apache is slightly faster (~4% or so) for >> files over 16KiB than Tomcat APR, and materially faster (~44% or >> more) than all other configurations of Tomcat (especially for larger >> files). > > Tomcat+APR is so close to httpd as to be in a dead heat as far as I'm > concerned. We'll know more once the larger-scale tests have been run. If > you graph these numbers (or read very carefully), you can see that > Coyote+APR outperforms httpd for two of the samples. > > Also, the APR connector without sendfile is basically the same as using > the "simple" Coyote connector. I suspect the same is true of the NIO > connector, though it uses a different strategy for reading and writing, > obviously. > > Something is obviously amiss with sendfile-enabled NIO connector, > though. Suggestions from those who know would be appreciated. > > - -chris > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v1.4.9 (MingW32) > Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ > > iEYEARECAAYFAkoRxCEACgkQ9CaO5/Lv0PBw1QCgs4g8fZk4ESSC7dDpVEZoAnah > HmQAoJk7FshdtZlboIG+niTRy0Lb5zRP > =6B0w > -----END PGP SIGNATURE----- > > --------------------------------------------------------------------- > To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org > For additional commands, e-mail: users-h...@tomcat.apache.org > > --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org For additional commands, e-mail: users-h...@tomcat.apache.org