[Bioclusters] requesting help for computational server setup (PE 4600)

Joe Landman bioclusters@bioinformatics.org
Wed, 17 Sep 2003 03:27:24 -0400

Hi Karthik:

On Wed, 2003-09-17 at 15:04, karthik viswanathan wrote:
> We have a PowerEdge 4600 server running RedHat 9. The system spec is given below.


> The main usage of this server is for bioinformatics computational work. Clients
> run search/match C or C++ programs on this server through protocols like ssh.
> The no of clients will be not more than 4. This server is also acting as a file
> server for the same clients. T Get 3 newhe performance is not satisfactory so we

I can't parse this "T Get 3 newhe performance is not satisfactory".  I
assume that you meant something like 

	"To get 3 new ..." 

and then I lost the meaning.  Could you elaborate a bit more?

> are planning to do some modification in the configuration. One of our plan is
> promote a client workstation(dell precision-340, 1.5 GHz & 1GB ram) to a file
> server and switch the two SCSI hard disk from the PowerEdge server to this
> workstation. Add 3 new SCSI hard disk (37 GB) to the PowerEdge server and set a
> RAID 0, if possible also add a second identical processor to the server.
> We are interested to know if this will improve the performance of the
> computational server. Also it would be helpful if you could suggest any other
> alternatives for improving the computational performance for bioinformatics
> work. If any of you have poweredge 4600 please share your experience and ur

Most important questions:

1) what software are you running, what type of inputs, what size data

2) have you run any of the standard system performance viewing tools?  

3) what are the contents of your /etc/raidtab if you are using a
software RAID0, and if not, what is the chunk size on the hardware

4) are you running NFS?  What is your network speed?  What are the

Initial observations:

1) ext3 is not as good as some of the other file systems for specific
work loads.  It might be worth looking into XFS

2) many bioinformatics tools are great with a divide and conquer
approach.  If we understood what tools you are running this might help
facilitate this discussion.

> system configuration. Suggestions on disk partitioning, kernel and file system
> to use will also be helpful

Lots of tuning possibilities, but we don't know where your systems hot
spots are yet.

> Thanks for your time and help
> Karthik
> PS: We are more interested to improve the computational power than file handling!

Sometimes the two are very much related.


Joseph Landman, Ph.D
Scalable Informatics LLC,
email: landman@scalableinformatics.com
web  : http://scalableinformatics.com
phone: +1 734 612 4615