On Sat, Aug 6, 2011 at 4:27 PM, Christian Paro <[email protected]> wrote:

> But is there any performance advantage to filling one VDisk before making
> use of the next (or, for that matter, in using multiple small VDisks rather
> than keeping things simple and using one large VDisk)? I would think that,
> since it's all backed by the same underlying store, it wouldn't really
> matter how you fill it so long as you're making sure you use that fast VDisk
> space before falling back to slower media.

The reason for doing multiple VDISKs for swap (other than the 2G
restriction) is to change the allocation strategy. Back in the early
days, my slides had the warning "Non-mainframe Linux people will
misunderstand and get it wrong" - apparently still true.

Doing parallel VDISKs to spread the I/O does not help you since it is
limited by CPU. To avoid fragmentation you define a series of VDISKs
for swap, each with different priority. The size of each is determined
by the memory requirements profile. Without this, swap to VDISK
basically makes no sense. Nor does it make sense to define swap space
on dedicated real disk.

Rob

----------------------------------------------------------------------
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
----------------------------------------------------------------------
For more information on Linux on System z, visit
http://wiki.linuxvm.org/

Reply via email to