Randy Patterson wrote:
So on a dedicated system that is used for nothing but running a processor
intensive application 24/7, do you think there would be any real increase in
performance from a custom kernel?
My gut reaction is "no," but I do not claim to have data to back that
up. If there is a measurable inefficiency introduced into system calls
to a loaded module versus a built-in module, then you could arguably get
some minimal throughput gain over a long time period given the situation
you're describing.
For actual data, you might try asking on the mailing list for the Debian
Kernel Team, or even at the Linux Kernel Mailing List. Please read some
FAQs first, because this may well be a question that has been
asked-and-answered to death.
Since you're admittedly using "older" systems anyway, we're talking
about (at the very most) losing a tiny percentage of efficiency on
machines that weren't going to be the major contributors of throughput
in the first place. I'm thinking I would be a Bad and Evil person if I
encouraged you to custom compile your own kernels for _only_ the purpose
you are describing. (If there were some educational goal on your own
part that you were pursuing, it would start to make more sense, IMHO.)
DW
--
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org