IUBio Biosequences .. Software .. Molbio soft .. Network News .. FTP

What's a good GCG server?

Tim Cutts tjrc1 at mole.bio.cam.ac.uk
Thu Sep 4 07:41:33 EST 1997

We recently bought an SGI Origin 200.  This is nicely expandable, on a
small scale (for $20000 I guess you'd initially get one with two
processors, assuming a decent quantity of RAM and disk space as
well).  It's upgradeable to a maximum of four processors.  To expand
further with an SGI solution you need to buy an Origin 2000, which is
probably out of your price range - it was certainly out of ours!

We are a similar size outfit to you; about 1200 users, 30 or so active
at any one time, mostly running pine.

Currently GCG is probably not running at peak efficiency since version
9.0 does not support IRIX 6.x, other than indirectly by support for
IRIX 5.3.  However, version 9.1 will support IRIX 6.4 directly, which
I expect will give a significant performance boost.

Our machine now has the full four processors, and I have *never* seen
it struggle, even running four fasta's and a couple of protein
structure simulations simultaneously.  Its disk access is also very
fast, due mainly to having four separate SCSI channels (although most
of our disks are on just two of them, the two Wide channels).

IRIX 6.4 also has some rather neat features for suspending processes,
rebooting the machine and then resuming the processes, which can be
very useful if you have long-running jobs but need to reboot the
machine for some reason.

On the downside, I have not been wildly impressed with SGI's quality
control, although they fixed our initial problems very promptly, and
I've had no further difficulties since June.

When we were considering this machine we also looked at Digital, and
were loaned a 2100/250.  It has to be said the Alpha is a faster
processor, but as I understand it the SGI has much better I/O
capability (the AlphaServer mentioned above had a single 32-bit PCI bus
capable of 135 MB/sec, whereas the Origin 200 has either one or two
64-bit PCI buses, and four separate SCSI channels, two at 20MB/sec and
two at 40MB/sec), which is quite a major consideration for database access
programs like fasta and blast.  The four CPU Origin 200 has a
sustained total I/O capability (allegedly) of over 2GB/sec although I
guess you'd never achieve this since even the RAM access bandwidth is
less than this.

I find IRIX easier to administer than Digital UNIX, but having said
that, IRIX has more security problems than Digital UNIX does, so you
have to watch out for those.

We briefly considered Sun, but the price of an Enterprise 3000 box
matching the performance of the Origin 200 was way too high.  Plus
Solaris 2.6 is still not here, which means that Sun boxes are still
limited to files of < 2.1 GB.  Given the speed the sequence databases
are growing, I'm a bit wary of that, even though most of the databases
are split into sections these days.

Just a few thoughts.


T J R Cutts                                           Tel: +44 1223 333596
Dept. of Biochemistry, Tennis Court Rd.,              Fax: +44 1223 766002
Cambridge, CB2 1QW, UK

More information about the Info-gcg mailing list

Send comments to us at biosci-help [At] net.bio.net