cache

quick question: of the performance data that you get from the outside world, what have you all seen to be the primary factors in execution time? for example, in the early stages of the binarycloud system design, I decided to just go with separate files for doing logically separate things, down to a fairly small granularity. it is my _belief_ that most of the actual execute time associated with binarycloud init process is directly related to the filesystem. that's mostly from some observations of the system in production: on IDE machines, times are ~.2 - .3 sec, on scsi machines with otherwise similar specs, load times kick down to ~.1 reliably. is that just something else? or is my assumption that this is directly related to the speed of your filesystem access correct? anyway, of the little I've read re: cache, I assume that it's going to cache on first run, and keep bytecode copies in memory? _a oh yes, if you're going to respond, please cc: me directly. :) -- Alex Black, Head Monkey enigma@turingstudio.com The Turing Studio, Inc. http://www.turingstudio.com vox+510.666.0074 fax+510.666.0093 Saul Zaentz Film Center 2600 Tenth St Suite 433 Berkeley, CA 94710-2522