CLMS Parallel Computing Cluster
What is it?
The Computational Linguistics program maintains a cluster of several servers (referred to as "compute nodes") that can run multiple processes in parallel, allowing large jobs -- if properly constructed -- to be run faster than they could on any single computer. We offer two high-performance computing frameworks:
- Condor can run any type of software, but leaves the details of handling the data to you. It's essentially a very flexible scheduler for matching batch jobs with available compute nodes.
- Hadoop can only run programs written using the MapReduce framework; however, it handles many of the intermediate data processing details itself.
Equipment for the cluster is partially funded by the Student Technology Fee.
How do I get an account?
If you do not yet have a Computational Linguistics account, fill out our online account request form
. You must have a valid UW NetID.
How do I access the cluster?
Open an SSH connection to patas.ling.washington.edu port 22 or dryas.ling.washington.edu port 22. You can also access files on the cluster directly via Samba filesystem shares. See HowToAccessPatas
for more information.
How do I use Condor to run my parallel jobs?
Visit the HowToUseCondor
Wiki page for a quick introduction. You may also want to view the PDF presentation
I wrote for orientation.
Software notes and reference information
General Tips and How-Tos
A listing of linguistics software on Patas can be found in the CompLing database
Patas? What's that?
are slender, ground-dwelling primates found in central Africa. They can run faster than any other primate, reaching speeds of up to 55 km/hr.
Topic revision: r36 - 2012-04-27 - 23:44:46 - brodbd