Howdy, If you''ve been on the IRC channel the last couple of days you may have seen some of my graphs showing session container performance vs number of sessions. Well, I killed a bit of time in the office today and wrote up a little document that gives a little more detail about the tests and the results. The document was thoroughly reviewed and mercilessly corrected by the notoriously language-anal Marcel Molina Jr. (aka "noradio"), and version 1 is now ready for public consumption. I welcome all discussion, suggestions, questions, and criticisms. The document can be viewed at: http://elitists.net/~kain/sessions/index.html Thanks -Scott _______________________________________________ Rails mailing list Rails-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org http://lists.rubyonrails.org/mailman/listinfo/rails
* Scott Barron (scott-HDQKq3lYuGDk1uMJSBkQmQ@public.gmane.org) [041124 14:33]:> The document can be viewed at: > http://elitists.net/~kain/sessions/index.htmlScott, Great work! I''ve been eyeing Michael Granger''s memcached client for Ruby as another alternative to session and object caching for use with Rails (and AR). I don''t know how easy it would be to set up a memcached session handler to try with your test suite, but if anyone''s interested in the Ruby memcached client (or memcached in general), there''s more information here: http://groups.google.com/groups?selm=8A80E684-3680-11D9-ADDC-000A959D1A74%40FaerieMUD.org Rick -- http://www.rickbradley.com MUPRN: 980 | Well in fact, I like random email haiku | Macs, Windows and Unix for | different reasons.
Really cool. Minor nitpick from another language-anal person: "For a baseline test, I did a run with Rails set to not keep session data." Better usage here would be "For a baseline test, I did a run with Rails set not to keep session data," or even better would be a rewording like "For a baseline test I did a run with sessioning turned off." Splitting an infinitive with "not" inbetween is ugly. Scott Barron wrote:>Howdy, > >If you''ve been on the IRC channel the last couple of days you may have >seen some of my graphs showing session container performance vs number >of sessions. Well, I killed a bit of time in the office today and wrote >up a little document that gives a little more detail about the tests and >the results. The document was thoroughly reviewed and mercilessly >corrected by the notoriously language-anal Marcel Molina Jr. (aka >"noradio"), and version 1 is now ready for public consumption. I >welcome all discussion, suggestions, questions, and criticisms. > >The document can be viewed at: >http://elitists.net/~kain/sessions/index.html > >Thanks >-Scott > > >------------------------------------------------------------------------ > >_______________________________________________ >Rails mailing list >Rails-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org >http://lists.rubyonrails.org/mailman/listinfo/rails > >
I have to say, after having read it, that you did a great job. It''s great to have some real-world top-notch analysis to demonstrate that rails could be a serious contender in the enterprise arena. Carl
On Wed, 24 Nov 2004 15:17:33 -0800, Carl Youngblood <carlwork-0CEYHQKyN7s@public.gmane.org> wrote:> I have to say, after having read it, that you did a great job. It''s > great to have some real-world top-notch analysis to demonstrate that > rails could be a serious contender in the enterprise arena.I agree, excellent work Scott! I think one of the interesting questions now is (and you''ve mentioned this yourself in the document) how different filesystems would affect the PStore performance. While I''m not looking to start a platform war here or really have any figures to back it up with, NetBSD is probably not the most widespread os for typical Rails applications. So I think it would be interesting to see some pretty graphs from different filesystems, like UFS (?) on FreeBSD, ReiserFS/ext3(/xfs) on Linux and HFS+ on OSX, if only to have some more data to look at. (Of course, if I get around to do this myself, I''ll send any findings to the list, but I''m a little low on time right now..) Maybe I''m misstaken here, but I also seem to remember something about how writing works differently in InnoDB vs. MyISAM in MySQL table types, I seem to remember one of them was locking the entire table (or db) when writing to a table and the other didn''t? But of course I can''t really remember any more details than that. And according to these graphs anyway the tabletype in mysql won''t really affect the session performance.. Again, great job Scott, thanks for sharing! -- johan -- Johan Sørensen Professional Futurist www.johansorensen.com