[Gllug] increase performance by increasing RAM or increasing in Core

[John Hearns]

2009/2/16 Peter Corlett <abuse at cabal.org.uk>:
>>
> Also, don't just limit yourself to /tmp. There are all sorts of other
> caches on a Linux box that you might not care about keeping over a
> reboot, and which are thus also candidates for a tmpfs. For example,
> Exim's retry and callout cache in /var/spool/exim4/db/.
>

Hell, why stop there?
Just run everything in a RAMdisk - then you'll have nothing slowing you down!

And no, I'm not being flippant. This is a common way of runing Beowulf
clusters - PXE boot into a RAMdisk,
and Bob's your Uncle. You can NFS mount /usr for all those pesky
binaries which users want to use.
You can even swap over iSCSI if you have no physical disks on the nodes.
(If you have disks you can use them for scratch local files and swap).
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug