FAQS for the CIPRES Cluster:

Q. How do I get an account on the CIPRES cluster?

A. Contact Mark Miller (mmiller@sdsc.edu)

Q. Am I eligible to receive an account on the cluster?

A. At present we are giving “accounts-on-request” to individuals within the CIPRES project. Thus means if you are on CIPRES funding, or are doing graduate/postdoctoral work for someone who is (see http://www.phylo.org/personnel/personnel.htm).

Q. How is time allocated on the CIPRES cluster?

A. We plan to allocate 1/3 of the available time on the cluster to CIPRES developers, 1/3 of the time to CIPRES analyses, and the balance to the community, giving priority first to the AToL projects.

Q. How is the CIPRES machine supported?

A. The CIPRES machine is located at the San Diego Supercomputer Center. We have systems administrators on staff, they are available 8-5 Pacific Time, and other times on best effort basis. They are easy to reach at cipres-support@sdsc.edu. This listserver will have archives that can searched for answers to questions.

Q. What kind of storage is available on the machine?

A. Local disk is available for storing or using data. Each node has its own local scratch disk which is good for this sort of thing. Directories on each machine for this use: can be created, e.g. /scratch/slocal/yourname. The storage quotas are very low in home dirs.

NFS Mounts many of the filesystems mounted on the CIPRES machines are nfs mounts, with central quota monitoring. You will, by default, have a small allocation here, maybe 300MB. Upon request we can add backed up NFS disk. Please use this option when necessary, but be aware we are charged for it, so don’t ask for it capriciously.

You can create a dir for yourself under /projects/cipres to work around the problem of low space allocation in home dirs. You need to be a member of the ciprescollabs group to create a dir there.

Q. What is the software environment?

The systems run the SDSC RedHat 4 Enterprise distro .
Additional Software packages:

Q. How are queing policies sustained?

In an effort to optimize the performance of the cluster for all our users, we have decided to implement a queing system. This will replace the current system where everyone can login in to any node, and deploy a task. The new system will be implemented on Monday, June 23, 2008. Please visit this page for information on SGE setup and use.

Q. What Phylogenetic tools are available?

The following tools are available:

PHYLIP, CLUSTALW/CLUSTALX, Blast2, ncbi-tools, FastdnaML, fastlink, tree-puzzle, readseq, molphy, treetool, njplot, loki, muscle, BioPerl, BioPython, MrBayes, Mesquite, Paup (All of these are 64-bit compatible Debian packages, which should also exist as RPMs -- and, if not, can be converted from .deb to .rpm format by standard tools.)

Paup is at /usr/local/apps/paup-4b10/bin/paup. RaxML 7.0.2 is in /projects/cipres/bin/raxmlHPC. Some other phylogenetics programs are in /projects/cipres/bi

Q. Can I port my serial application to the CIPRES resource?

Parallelization of serial code is a time consuming process. If you have a serial code and wish to parallelize it for use on the CIPRES cluster. Here are the steps to proceed through.

  1. First, make sure that your serial code is as good as can be and then use it enough to document its practical significance and its computational demands.
  2. In so doing, you will have created a presentation (talk or poster) That can be presented to someone who can be actually helpful in making the determination.
  3. If the code turns out to be both significant and computationally too demanding for a single processor machine, then, and only then, should he even consider parallelizing it.

At that point, you should consult with CS faculty at your home institution, and/or with David Bader or Bernard Moret to ascertain whether parallelization is a reasonable mode of attack. That should not be hard to do, because you already have had a poster and a presentation on your serial code at a professional meeting.

If the answer from the experts is "yes, it would help", then, and only then, should you start investing time into the parallelization effort. At that point too, you can also solicit help from CIPRES students or professionals;