Dave wrote: > Another example--I have a CLEARLY single-processor system. So why is > irqbalance activated in runlevels 3-5? I can grudgingly accept why it was > installed--it's easier to have a single kernel-utils package than have a > single processor and multi-processor versions. And apparently it exits > quietly when it discovers it's not needed. But why is it turned on in the > first place? I'm not sure that it's as clear as all that. Apart from laptops, most computers sold today could, in theory, be upgraded to need an SMP kernel. Anything P4-based could be upgraded to a Hyper-Threading CPU. Both AMD and Intel have multiple-core processors coming: the AMD ones, apparently, might not even need a BIOS update to work in some current motherboards. And, of course, you can always transfer the hard drive to something that does do SMP. Many servers are supplied with a dual CPU motherboard, but a single processor. Then (the rationale goes) the purchaser can upgrade later. This market is likely to buy Red Hat Enterprise Linux, or variants thereof. Given that irqbalance is installed anyway, and exits fast when it discovers that it isn't needed, it's probably better just to keep it there. You would, of course, need to install an SMP kernel. And you could say that I'm arguing for having one of those installed by default. I can see some merit to that argument, at least in RHEL. But it's the sort of install that people remember to do (if you add a CPU, and find the OS doesn't detect it, you tend to ask why). Activating irqbalance isn't like that. But then I'm probably biased. This computer is a dual-Celeron 433 system. James. -- E-mail address: james | I learnt the rules of rugby. There is only one rule. @westexe.demon.co.uk | "Skip it by any means necessary".
