[YLUG] Server purchase

[Roger Leigh]

"Richard G. Clegg" <richard at richardclegg.org> writes:

> I wonder if I could get the advice of the massed ranks of the group
> for two things here.  I'm trying to get a high-availability server
> going within the dept of Mathematics here at York.  We've had lots of
> reliability issues associated with hardware failure recently (three
> downtime incidents in four weeks after a few years of trouble free
> running -- three independent hardware failures).
>
> Anything I need to think about?

Have you looked at http://www.linux-ha.org/ ?


For really good fault tolerance, you might be better with hardware
which is better capable of dealing with failure.

Have you considered POWER-based hardware, or are you tied to ix86 for
the software the server runs?

Systems like IBM's OpenPower 720 are very expensive, but would be
extremely reliable.  However, they now do systems like the p5 series
(e.g. p5 505 Express), a 1U rackmount, which looks like it meets all
your specifications; It's also quite reasonably priced for IBM.  They
will both run Debian just like on Intel.

http://www-03.ibm.com/systems/p/hardware/express.html

(I've not used one myself, but have seen them running.  These are the
first IBM kit not to require a separate HMC for LPAR.)


It might also be worth looking at Xen, which can migrate Xen instances
to backup nodes in the case of failure.  Having seen this demonstrated
by the Xen author, I can attest that this is really impressive.


BTW, hardware RAID is often slower than software RAID, because with
software RAID you can split the disks so that each one can saturate a
single IDE/SATA/SCSI controller/bus.  With hardware RAID, you often
have a bottleneck in the RAID controller.  I've been to a talk by one
of the Linux SW RAID guys, showing that its performance can increase
linearly with each new controller you add.  OTOH, software RAID may
potentially be less reliable, and (depending on the hardware) may not
be hot-swappable.


Also, for extremely fault tolerant filesystems, Sistina/RedHat's GFS
(Global File System) clustered filesystem might also be something to
investigate.  This gives you a filesystem distributed over all the
nodes in the network, with no central server.  It can cope with nodes
failing, and you can use RAID or normal disks on each node in the
cluster.  It can be thought of as RAID over an entire cluster of
networked systems, but it's more than that.


Regards,
Roger

-- 
Roger Leigh
                Printing on GNU/Linux?  http://gutenprint.sourceforge.net/
                Debian GNU/Linux        http://www.debian.org/
                GPG Public Key: 0x25BFB848.  Please sign and encrypt your mail.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 188 bytes
Desc: not available
Url : http://mailman.lug.org.uk/pipermail/york/attachments/20060519/f49aefbd/attachment.bin