On Mon, 21 Jun 1999, David van der Spoel wrote:

> Hi,
> 
> I considuer buying a large server case which should run Linux off a single
> (small) SCSI disk+controller and serve 8  9 Gb disks from 2 extra
> SCSI controllers. I intend to run RAID 5 on the 8 disks, with one or
> two spare disks.
> 
> Questions: 
> 1. Will the second spare disk increase redundancy such that if two
>    disks break, the data will still be safe?

It ought to, but I always wondered about this... The disk is sitting
there, spinning but the head not moving, waiting to take over from a
failed drive. Is this any better than it sitting on a shelf? What actually
fails. The electronics? (in which case it has the same chance of failure
as any of the other active drives), the motor, a head crash? It all sounds
a bit dubious to me. I guess what you gain is convenience more than
anything else though.

> 2. Will the case melt down because of heat?

Depends on the case! Make sure your PSU's can handle it. It might be
better to get separate cases for the processor and drives.  I'm currently
using cases from Acme Technology (http://www.acme-technology.co.uk/). They
sell removable caddys from drives as well as redundant PSU cases and so on
- both stand-alone tower types and 19" rack mounting types. You should be
able to find someone local you your area though, but their web site might
give you an idea of what sort of cases are available.

I feel a good case is worth it for a server - afterall, I don't know what
data you have, but the data I have to look after is quite valuable to the
company I work for!

> 3. Should I, in light of coming PIII patches get a mobo with PIIIs?

I'm running a dual Xeon III motherboard. I actually had real problems
getting PIIs! I was told by my usual suppliers that they couldn't get
them anymore, so had to get PIIIs.

> 4. I already have the disks and scsi controllers (Fujitsu/Q-logic), but
>    is there anything otherwise that will increase performance as an NFS
>    server? 

Large CPU Cache, lots of fast memory, fast (100Mbs) Ethernet and a
switched Ethernet network. I only have 512KB cache Xeon's, but I obviously
have the option to upgrade to larger cache Xeons if I feel the need to. It
boots from an IDE drive and I have a 2nd identical IDE drive on the 2nd
controller which I need to look into making a mirror of the first drive.
The data drives are 4 x IBM 18GB drive, 2 on each controller (MB has
onboard dual channel Adaptec beast) I also have a 35/70GB DLT tape drive
to back it all up! Remember that RAID isn't a substitute for backup &
archive...

My server has 512MB of ECC RAM and an on-board Intel etherexpress chip. We
have a fully switched network and all the switches are capable of 10/100
full duplex operation. My main server serves about 100 PC's running mostly
NT and a dozen sparcs running Solaris and a small handfull of other Linux
machines. (Although I'm seeing some NFS problems with the sparcs which I
need to look into )-:

Good luck!

Gordon

Reply via email to