RE: [CFCDev] CFC Performance question

[Nando]

OK, it would be interesting to see what anyone has to say on this. It seems to report the timings correctly on calls to pages that don't employ cfc's - but i would definitely turn in off on your dev machine to check page execution timings in the cfc era. it *might* just be (if we are lucky!) the debugging output that's hanging up the processing. i wouldn't be surprised, as there is already some evidence in this direction, and the debugging output for cfc's certainly looks complex enough to slow anyone down.   -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]On Behalf Of Jim Davis Sent: Monday, November 10, 2003 1:16 PM To: [EMAIL PROTECTED] Subject: RE: [CFCDev] CFC Performance question I have my doubts about the quality of the debug reporting as well – but in my case SOMETHING is eating the time.  Most of the time a process takes, say 20 seconds – then, for no apparent reason, 2 minutes.  The debug output attributes it (usually) to a single method call.  Now, of course, that method probably isn’t at fault – it’s something that just happened while that method was in process.  But (in my case) the page does take about 2 minutes to load, then, next hit – back to normal.   I’ll be seeing Mike Nimer on Wednesday – I’m pretty sure he wrote the standard Debug page – I’ll ask him about it.  It’s actually an “open” unencrypted page – we could look at it (and modify it) if we like.  I just never seem to have the time to dig into it.   The caching tool looks interesting – I may very apply it, thanks.   Jim Davis     -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Nando Sent: Monday, November 10, 2003 4:07 AM To: [EMAIL PROTECTED] Subject: RE: [CFCDev] CFC Performance question   Jim,   I've seen these too (weird performance dips) on occasion as you describe. A couple of things i stumbled across in the last weeks ...   One is that for some reason, debug mode reports timings seem to be very inaccurate when using cfcs - in my quick tests, the more you employ during a request, the more inflated the reported timings become. Ben Edwards pointed this out to me when i asked about why the Mach II framework was taking so long. When i ran the same pages with debug mode off, using getTickCount() only around index.cfm - the reported timing was 5 times less! Maybe someone else has more accurate information on this descrepency.   The other is a very cool caching tag that Brandon Purcell posted on his site.   http://www.bpurcell.org/blog/index.cfm?mode=entry&entry=963   He states that you can get upwards of 700 pages a second under full load testing with this thing. And it's VERY simple to use. Surprise surprise - for once it came easy! Each page is cached individually as well, so you don't get a big hit that one user has to take. Even then however there are odd weird performance dips: a method that takes 0ms 1,000 times will be reported as taking 60 seconds once.  There’s no specific load involved (in fact the machine is barely sweating) and no outward sign of trouble – but it just hangs (seemingly, I’m guessing waiting for some timeout someplace).   I’d like to track down why that’s happening… but it’s so inscrutable I doubt I’ll find it…