There was a posting on the CakePHP mailing list today from someone asking for sources of advice on "starting big applications". Having gone through this at my previous job, I posted the following response:
Just to inject some sanity into what should be a very interesting thread, I urge anyone building an application to simply BUILD IT and worry about the profiling / benchmarking for AFTER you've built it. Trust me when I say this: no application of any significant size survives long enough to be scaled past it's original intent without wholesale architectural changes. You *WILL* be rewriting things, you *WILL* be rethinking how your application is organized. Like Death and taxes, it is inevitable. But you can worry about it later. I say this as someone who has been through the "let's build something massive" process, only to see it fail horribly due to all sorts of preconceptions at design time that resulted in major rewrites that no benchmarking would've helped. We didn't know what the bottlenecks were until we actually saw them get hit, and then we set about rewriting the bottlenecks. A little slipshod for sure, but there were many things that happened that we could not have anticipated when we were building it. If your site is so busy that you need to do benchmarking and other refactoring work, then you have accomplished more than 99% of other websites out there and deserve a round of applause. Until then, Just Build Something, Damnit and worry about fixing any performance issues later. Before everyone jumps on me, please understand that I am saying this out of experience. Yes, there are certain things you can do in terms of building a cache-friendly app, or configuring a web server for optimal use but those are so application-specific I hesitate to recommend those things until you actually know what you need. Premature optimization lies on the path to developer madness. You won't know what challenges your application will face until people other than you are actually using it. ;) Hope that helps.
My new mantra for programming in 2007 is Just Build It, Damnit! I've been procrastinating too much on several personal projects so it's time to put my money where my mouth is.