> And that mindset is how you wind up with excessive server sprawl. In almost > every case, a single application running on distributed hardware is going to > look cheaper than running on Linux on System z. The question is, which > part(s) of the server farm _could_ be run on z/VM or some other > virtualization platform, would they meet their business objectives in that > environment, and would there be sufficient cost savings to justify the more > expensive hardware?
The question is when does an extra 1U PC box become cheaper. I can put 128GB into a pair of cheap x86 boards. Thats a lot of memcached space and I wouldn't like to imagine the price of that on a z/VM box. You can certainly plot ownership cost curves including licensing, power, etc for your workload set and I don't doubt some of it looks better on each - especially if you are required to have the big box for another workload so it's a sunk cost anyway. Another thing not touched on much yet of course is reliability. It's easy enough do to high reliability with bulk PC hardware but it's a slightly different cost model (especially with a low number of systems) to doing typical 'internet grade' reliable. > Not every workload is a good candidate for running virtualized, whether on > z/VM or something else. Not every workload is a good candidate for System z, > virtualized or not. As I said previously, picking the right tool for the job > is more complex than most people want to have to deal with. But if you're > not going to paint yourself into a corner, the bigger picture needs to be > considered for every workload that gets deployed. Indeed - but the bigger picture for z/VM here is probably bigger than the one being asked. memcached is a huge win because PC memory is cheap to buy (under $4 a gig) but while the I/O subsystem on a typical PC class machine is extremely cheap ($0.05 a gig), it compared to the processor is very very slow. So the question when deploying on a box with very fast I/O is IMHO not "should I put memcached on that box" but "do I need memcached on that box" - and also if the cache can be a lot smaller as a result. Alan ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 ---------------------------------------------------------------------- For more information on Linux on System z, visit http://wiki.linuxvm.org/
