[Gllug] Server setup

[Nix]

On Tue, 14 Dec 2004, Jan Kokoska stipulated:
> "There has been a lot of discussion about whether swap was stable on
> RAID devices. This is a continuing debate, because it depends highly on
> other aspects of the kernel as well. As of this writing, it seems that
> swapping on RAID should be perfectly stable, you should however
> stress-test the system yourself until you are satisfied with the
> stability."

I guess that RAID writes don't need to allocate memory, then. (Viz
latest FUSE threads and latent instability problems with all userspace
filesystems being used as an argument to keep them out of the kernel;
then others point out that exactly the same arguments can be used
against all network filesystems because of the need to allocate skbuffs
when there may be no memory left...)

> Industry practise in a not-so-small ISP I worked for, I don't care what
> you use in your home system and what has worked for you since summer or
> whatever.

If I had three disks in one of my machines, I'd be using RAID. (With LVM
on top of it of course.)

But all of my machines have two. No real point RAIDing that unless you
like losing half your disk capacity or reducing your reliability even
more (I/O performance doesn't count for me: my machines are rarely
disk-bound).

> Sorry for being impolite, though, it's just that I have people jumping
> at me "Xen virtualization solves everything"

Cool! Can a Xen box solve the Halting Problem? What about this 500-node
TSP: you have five minutes.

(No? Damn.)

>                                              and "UML is not inefficient
> anymore with the latest patches"

I'd phrase that as `UML is *much less* inefficient with the sysemu
patches'. (Specifically there are half as many context switches as there
used to be. Performance is now close to that of the host: precisely how
close depends on the number of syscalls you make, as syscalls are still
much slower than on the host. They're just much less much slower than
they used to be. :) )

>                                  and "software suspend is stable now"

Um. *cough* *splutter*

> personal systems but would not deploy it on massive scale where
> supposedly my job depends on it.

Well, people *are* deploying UML on massive scales and running UML
hosting services and the like. :)

(UML itself hasn't crashed on me since April this year, 2.0.24-1
or thereabouts. IT has a nasty network-freeze bug right now, but
that doesn't *crash* it, just makes it not talk to anyone for
a minute at a time every so often.)

-- 
`The sword we forged has turned upon us
 Only now, at the end of all things do we see
 The lamp-bearer dies; only the lamp burns on.'
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug