Yeah, of course everything is variable depending on the complexity of the CFC, etc. I've run some simple loop-based tests, which I know aren't very useful but it's a quick and easy way to get a probable idea of what's going on. I can instantiate 10,000 fairly simple beans in under 10,000 ms, which is less than a ms per instantiation. Which to me seems pretty freakin' fast. (this is CF8).
On 10/5/07, Barney Boisvert <[EMAIL PROTECTED]> wrote: > > What does "23 times faster" mean? That per-instantiation, it takes > 1/23rd of the time on CF8 that it did on CF7 (i.e. a 95% reduction)? > If so, that's a great improvement, but it's nearly devoid of meaning > here. What if CF took 1 second per instance on CF7? Then CF8 would > be expected to take 40ms per instance. That's hardly exemplary, even > though it's a huge improvement. It's also worth mentioning that > Adobe's going to publish the biggest numbers they can - real world use > is likely to be less. > > The specific case that I have tested was using LCDS. If I'm managing > a collection and I'm using CFCs as the server-side representation > (i.e. what the DAO that backs the assembler returns an array of), it's > slow. If the DAO returns structs with the "__type__" property set, > it's fast. Granted, using CFCs is WAY faster on CF8 than it is on > CF7, but it's still not fast enough, because you have to pay that > instantiation cost up front, even if they don't get sent client-side > immediately. In short, the server-side cost is greater than the > transmit cost, which totally negates the utility of CFC-based LCDS > managed collections. > > So yes, I've tested it, and the results pretty clearly indicated two > facts: > > - CF8 creates CFCs enormously faster than CF7 > - CF8 still creates CFCs too slowly to be instantiating more than a > handful per request > > cheers, > barneyb > > On 10/5/07, Brian Kotek <[EMAIL PROTECTED]> wrote: > > Just curious Barney, have you tested this? The numbers indicate that CF8 > is > > 23 TIMES faster than CF7 creating CFC instances. That is a pretty > gigantic > > increase. It seems to me that unless you're creating huge numbers > > (thousands+) of instances, the option to create true arrays of objects > > (instead of queries, for example) is actually a valid option now. > Thoughts? > > > > On 10/5/07, Barney Boisvert <[EMAIL PROTECTED]> wrote: > > > > > > If I have a dragster that runs the quarter mile in 10 seconds (really > > > slow), and I improve it to 7 seconds (a 30% decrease), it's still not > > > competitive. I'll grant you that CF8's CFC instantiation is much > > > faster than CF7's, but it's still far from sufficiently performant for > > > a lot of scenarios. > > > > > > cheers, > > > barneyb > > > > > > > > > > > > > > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~| Check out the new features and enhancements in the latest product release - download the "What's New PDF" now http://download.macromedia.com/pub/labs/coldfusion/cf8_beta_whatsnew_052907.pdf Archive: http://www.houseoffusion.com/groups/CF-Talk/message.cfm/messageid:290401 Subscription: http://www.houseoffusion.com/groups/CF-Talk/subscribe.cfm Unsubscribe: http://www.houseoffusion.com/cf_lists/unsubscribe.cfm?user=89.70.4
