I definitely think your problem is something aside from disk access. Have you monitored the reads it's doing under load and confirmed that it's really at it's max? Any decent filesystem will do in-memory caching of commonly requested blocks, so unless you're hitting a wide variety of disk-based content, you shouldn't be hitting the disks at all.
Do you have trusted cache enabled in CF and your CF templates on the same physical device as the graphics in question? If you do, then it's definitely not a disk issue. Bottom line, I'd make sure that's really your problem before you bury your CF server in a massive number of requests, and completely slaughter it's ability to serve real pages. cheers, barneyb On 10/6/05, Nando <[EMAIL PROTECTED]> wrote: > So i should move the cfcontent out of the getImage method, and return a > struct instead with the image type and image data. (or do i really need > different type attribute values for the different image types?) It's working > with the cfoutput tags, but maybe that's not as efficient as it could be. > > The reason i'm trying to serve these images from memory is because the > interface is very rich in graphics - and it seems the performance bottleneck > might be at the disk access point. There's plenty of memory and enough cpu > available, and when it bogs down, it's the graphics that are coming in one by > one, 15, 20 seconds apart. My reasoning is that all threads could have more > or less instant read access to memory, but they'd need to que for access to > the disk. Does that make sense? > > thanks for your help! > Nando > -- Barney Boisvert [EMAIL PROTECTED] 360.319.6145 http://www.barneyb.com/ Got Gmail? I have 100 invites.
