> That surprises me. We aren't a Linux shop--yet--and I'm not sure I'm > willing to start setting up Linux boxes will-nilly for this. I'm curious > why Intel would be better than sparc, though.
Try it out on a [higher-end] Desktop PC -- the performance may change your mind. Sun is hyping Solaris 10 through the roof for a reason, you know -- lots of mind/marketshare loss to Linux because of Solaris on Sparc's cost and lackluster performance. And note that Sun ships a lot of Opteron servers. Kinda makes you wonder, right? For running SM, I would expect a well-implemented PIII/1Ghz to perform about 3x faster than an UltraSparc III/400. > As for the load testing software, it just runs a script--say, login, > check mail, go to a folder, logout. Rinse, repeat. Once SM failed for > real world users, I needed some way to test it without using human > subjects. :) That's probably a good test, then. That whole process should take roughly 3-5 seconds, I reckon, depending on the size of the folder you're opening. > We use iPlanet for mail, and again, I can't see changing one component > of the system for this. In fact, our mail set up is pretty complex now. > Adding a different imap would make that worse. What I see in testing is > that the load goes way up on the initial login, then levels out. I > assume that's the sorting problem. What sort of web mail did you hope to offer? If they're all just speaking IMAP, won't these performance problems be universal? As for the load leveling off after login, well, that could be something different. Since the IMAP end of the pie takes place at every "click" in the application, I think the only things that are really different during the initial log in are: a) A new PHP session is created. Disk I/O problems? How are you storing sessions? By default, all php sessions are stored in a single directory. A Sun? Running UFS? Baaad disk I/O, particularly with lots of small files in a single directory. Consider veritas fs if you're going to stick with filesystem-based session storage, or use a different session backend such as mysql or postgresql (the latter being more preformant). b) Does your load tester pull down the pages as a normal web browser would, or does it take shortcuts? During login, the user is sent to webmail.php, which is just a frameset that turns them to two other pages. That makes the "login process" take longer than, say, a single folder load. I don't think you're necessarily dealing with limitations in SM, but with limitations that are stacking up throughout the app stack -- PHP's not the fastest, iplanet isn't doing you any favors, and your Sparcs aren't the fastest CPU's. At any rate, you'll definitely want a PHP accelerator and I think you could reap some benefits by running the web side on Linux/Intel. If nothing else, the ability to scale this thing horizontally on the cheap is attractive. There's more to this, but that ought to be a good starting point. John -- John Madden UNIX Systems Engineer Ivy Tech State College [EMAIL PROTECTED] ------------------------------------------------------- This SF.Net email is sponsored by: InterSystems CACHE FREE OODBMS DOWNLOAD - A multidimensional database that combines robust object and relational technologies, making it a perfect match for Java, C++,COM, XML, ODBC and JDBC. www.intersystems.com/match8 -- squirrelmail-users mailing list Posting Guidelines: http://squirrelmail.org/wiki/wiki.php?MailingListPostingGuidelines List Address: [EMAIL PROTECTED] List Archives: http://news.gmane.org/thread.php?group=gmane.mail.squirrelmail.user List Archives: http://sourceforge.net/mailarchive/forum.php?forum_id)95 List Info: https://lists.sourceforge.net/lists/listinfo/squirrelmail-users
