Would it be something as simple as a response time test to decide which
swap algorithm to use? In a perfect world I'd like to simply give Linux
one chunk of v-disk, and let the OS figure out how to use it. As server
apps grow it gets riskier to mess with the fstab to add more v-disks,
and we also end up with a bunch of non-standard disk layouts unless we
plan well in advance.

In the mean time, I think I'll standardize on 3 progressively sized
v-disk addresses starting at .5 RAM, with no DASD swap.

Thanks.


Ray Mrohs
U.S. Department of Justice
202-307-6896


> It would need to be something that works for all platforms, not just
> for our virtual disks.
> The idea is that by using fresh slots you have a better chance to
> build long chains of adjacent blocks in a single I/O operation. If
> utilization is low enough, the "moving cursor" is an easy way to do
> that.
> With vdisk we're willing to take the overhead of many small I/O
> operations because the device is already fast enough (especially when
> your driver can exploit synchronous I/O completion).

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to