Migrating to Unix: SGI vs. SUN

Tim Cutts tjrc1 at mole.bio.cam.ac.uk
Fri Sep 13 02:17:31 EST 1996

In article <1996Sep12.155933 at bphvax>,
Charles A. Alexander <caal at bphvax.biophysics.rochester.edu> wrote:
>We're presently evaluating the pros and cons of migrating our GCG 
>package to a UNIX platform.  It seems like a lot of the freeware 
>biocomputing apps are being written for UNIX and though I love my VAX 
>and VMS System Management, it seems like I would be doing my users a 
>disservice by not providing at least the option of other software.
>We also want to provide molecular modelling tools for the people we 
>support and it seems that SGI and SUNS have a lot of tools available for 
>these applications as well.
>Anybody have any recommendations on what kind of machines would be 
>able to handle the kind of processing power required for both these 
>applications for a peak load of about 10-15 users simultaneously.  We 
>have a total of about 50 users and growing.

mole.bio.cam.ac.uk is a three year-old SGI Indigo2 with a 150MHz R4400
and 128Mb RAM.  Not a stupendous machine at all, and we are looking to
replace it soon.  However, we are trying to do what you are doing, but
on a rather larger scale;  mole has 1,500 users, with usually 30-40
logged in at any one time during the day.  FASTA searches are getting
a bit slow these days, which is why we want a more powerful machine.
However, it seems that an SGI workstation would probably suit you
quite well.  However, *don't* to molecular visualisation on the same
machine; you will kill the interactive performance of the remote
users.  Our protein modelling group uses SGI Indigo2 machines as well
for their NMR work, and are very happy with them.

I can't speak for Sun machines really, but it seems that the majority
of the bioinformatics world are either using SGI or DEC.  We have a
Digital Alpha machine on loan at the moment for evaluation.  A 250MHz
ev5 Alpha processor seems to be about on a par with a 195 MHz R10000,
although the R10000 might have a slight edge in our tests.

Having said that, so far I prefer IRIX to DEC Unix.  IRIX seems to
have more flexible process sheduling, and this is very important if
you buy a single processor machine which is also going to handle
interactive tasks such as e-mail (which ours does).  With
multi-processor machines this is less of a problem because you can
restrict certain functions to certain processors.

>I don't want to start a holy war but should we look at Suns or SGIs ?  

You should definitely consider DEC as well.  It also gives you the
flexibility to run OpenVMS if your users (or you) decide they loathe
UNIX :-)  Or even Windows NT; a fair bit of software seems to be being
ported to it these days.

>Anybody care to share their experiences ?  How stable are the OS' and 
>how difficult are they to manage.

We used to use VAXen, although this was before I joined.  In many ways
UNIX is much more messy than VMS (according to my boss/predecessor),
and it's probably less secure.  However, IRIX is pretty stable (as
long as you use a 'standard' version, which at the moment means 5.3 or
6.2).  mole runs quite happily for months without a reboot.  I only
had to shut it down last week because a hard disk failed (joy!).

>My limited experience tells me that 
>SGIs are easier to handle.  I'll apologize in advance if my ignorance 
>is showing.

Well, I know very little about VMS, so I can't really compare.

>Thanks in advance. BTW, anybody have an updated spec sheet on how GCG 
>performs on various platforms ?

Don't trust these things.  The comparisons are always totally
ludicrous.  In our tests so far, on a Dec AlphaServer 2100 5/250 and
an SGI Indigo2 with a 195MHz R10000 processor, both machines are about
3x as fast as the 150MHz R4400 doing either GCG or Pearson FASTA
searches.  From the SGI data I have seen, performance scales rather better
with multiple processors on SGIs than on DECs, but you'd expect SGI
benchmarks to say that, of course.


Dept. of Biochemistry                                 Tel: +44 1223 333596
Tennis Court Rd.
Cambridge, CB2 1QW, UK

More information about the Info-gcg mailing list

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