From UConn Bioinformatics Service

Welcome to the University of Connecticut Bioinformatics Facility Wiki help page

(For any questions regarding problems or help request : pascal.lapierre@uconn.edu[1])

Please acknowledge us!

If you use the cluster in the process of writing a paper, please insert wording like this in your Acknowledgements section:

"The UConn Bioinformatics Facility provided computing resources for the Bayesian and maximum likelihood phylogenetic analyses performed for this study."

Also, when the paper is published, please remember to tell Pascal about it so that we can include it in our annual report.

Important Request!

Please do not run jobs outside of the queue systems! If you need to run an interactive job then please use qrsh or lslogin to have the queue resource manager (either SGE or LSF) assign you to an available node where you can then run the job. There should be no need to bypass the queue systems under normal circumstances. It is important that you follow this protocol so that we can keep an accounting of the use of the clusters in order to ultimately justify their continued support by the University.

In addition, please do not put jobs into the background from a "qrsh" or "lslogin" session. This also puts those jobs outside the queue system and thus we lose the ability to track them for accounting purposes. "qrsh" and "lslogin" sessions are meant for interactive sessions, using the queue system, where a task is completed during the session and then the session is ended. Any long term jobs should be submitted using "qsub" or "bsub" Thank you for your help.

Disk usage on the clusters and disk quotas on the Dell Linux cluster

The Bioinformatics Facility clusters were not designed to provide long term general storage for users. Please offload or delete any files that are no longer being used and make sure that you have copied any important files back to your local machines. You can check how much disk space you are currently using on the Dell cluster by running the command "quota". The quotas are reported in the number of 1 KByte blocks.

Getting started on the Xserve cluster

Requesting a user account
iNquiry Web Access
Secure Shell and SFTP or scp transfers
Getting on a node (qrsh and ssh)
Running short jobs (under 6 hours)
Basic Unix commands and programs available on the cluster
Running jobs without using the qsub command
Using the qsub and qstat commands
Stopping jobs using kill and qdel

Getting started on the Dell Linux cluster

Requesting a user account
Secure Shell and SFTP or scp transfers
Getting on a node (qrsh, lslogin, and ssh)
Basic Unix commands and programs available on the cluster
Running jobs without using the bsub command
Using the bsub, bjobs, and qstat commands
Running an mpi job using bsub
Running a multi-threaded job using bsub
Stopping jobs using bkill
A quick reference guide to useful LSF commands can be found here.

RaxML on Dell Linux cluster

Quick tutorial on how to run raxmlHPC on the Dell cluster

Multi-threaded Applications on the Linux Cluster

Qsub parameters needed to run multi-threaded applications correctly can be found here.

Current News

Queue Submission for Dell Cluster
The Dell Cluster uses LSF (Load Sharing Facility) as its resource management software. LSF uses a different queue submission command and syntax than SGE (used on the Apple Xserve cluster), which uses qsub. Please review the use of the bsub command prior to submitting jobs on the new cluster.

New version of clustalw :: Mar. 18, 2008
The latest version of clustalw from EMBL-EBI, called clustalw2, has been installed on both clusters. The command "clustalw2" will run the latest version, and on the Xserve cluster the old version can be run using the command "clustalw". You can take a look at the EMBL-EBI web site to check out what is new and/or different in their version.
Note: clustalw2 appears to run slightly slower than clustalw.

Blastology 101

See Blastall help page for information on Blastall commands and genome comparisons page for basics about finding homologs, paralogs and orthologs across genomes.

Running the parallel version of clustalw on the Apple Xserve Cluster

Before using mpi with any of the parallel programs you'll need to set up passwordless SSH. If you don't yet have this set up then the easiest way to do this is to enter the following three commands from your home directory:

  ssh-keygen -t dsa -f ~.ssh/id_dsa -N ""
  
  cat ~/.ssh/id_dsa.pub >> ~/.ssh/authorized_keys
  
  chmod 644 ~/.ssh/authorized_keys

See the clustalw-mpi help page for information on running the MPI-enabled version of clustalw.

Running the parallel version of MrBayes on the Apple Xserve Cluster

Before using mpi with any of the parallel programs you'll need to set up passwordless SSH. If you don't yet have this set up then the easiest way to do this is to enter the following three commands from your home directory:

  ssh-keygen -t dsa -f ~.ssh/id_dsa -N ""
  
  cat ~/.ssh/id_dsa.pub >> ~/.ssh/authorized_keys
  
  chmod 644 ~/.ssh/authorized_keys

See the MrBayes help page for information on running the MPI-enabled version of MrBayes.

Community Rules to keep in mind when using the Cluster

There are no official limits on the number of jobs you can submit, it all depends on the overall load of the system. During period of low activity, you are free to use more nodes than when many other users are running analyses as long as they are not long running jobs, e.g. less than a day or two running time. To have an idea of the load on the Apple Xserve cluster, you can either use the qstat command, look at the QueueStatus on the iNquiry portal, or go to here for a web overview of the Xserve cluster usage. For the Dell Linux cluster you can go to here.

NEVER run anything on the head node of a cluster (default login node). To log onto an idle node, use the qrsh or lslogin command.

Always keep note of what and where you are running your analyses. If something goes wrong, you can always log back on the node where your analysis is running (ssh node0## for the Xserve cluster, or ssh compute-1-# for the Linux cluster)) and terminate the process (kill).

Main Page