I/O Scheduler Framework

Justin C. Sherrill justin at shiningsilence.com
Sat Feb 28 09:48:00 PST 2009


On Sat, February 28, 2009 5:35 am, Alex Hornung wrote:

> - Dividing policies into first order (fair share, anticipatory, ...) and
> second order (bandwidth limiter, ...)

Should we establish how many policies we'd actually need?  The first-order
items described seem mutually exclusive, and a bandwidth limiter (which is
a good idea) is the only one so far at the second order.

I'm thinking that what Simon suggested about just having a single
scheduler may work just fine because there aren't enough modular items to
really require the extra work of extra scheduler 'layers'.






More information about the Kernel mailing list