This is the home page for the High Performance Computing space.
Education and Training
Every year the Mechanical Engineering (ME) department, the Section on Statistical Genetics (SSG), UAB IT, and others at UAB conduct a High Performance Computing (HPC) "bootcamp" covering many topics of interest including: intro to Linux, intro to job queuing, setting up job scripts, intro to grid computing. This bootcamp has a specific focus on education and training for the resources available here at UAB (outlined below).
The handouts, agenda, and videos for past bootcamps can be found at http://www.soph.uab.edu/ssg/courses/hpcbootcamp.
Contact Vinodh Srinivasa (firstname.lastname@example.org), Sr. IT Manager, SOPH for more information.
UAB Beowulf clusters (home page)
The most up-to-date information regarding these Beowulf clusters is available on the ME wiki so we don't go into much detail here except to say that these resources are appropriate when you plan to run many hundreds, thousands, or more jobs that require CPU processing, but not much RAM.
This guidance applies to many of the jobs that the folks at the SSG are interested in and run (think simulation study with a single algorithm and many independent, simulated data sets or a single algorithm and single data set with many different combination's of input parameters). In other words, the SSG currently makes the most heavy use of the Beowulf clusters compared to the other HPC resources described on this page.
New environment variables have been added to the UAB Beowulf clusters(currently Cheaha only) which point to the scratch space. More information on the usage policy and variable names is available here -> http://me.eng.uab.edu/wiki/index.php?title=Cheaha#Scratch
Local Linux servers (home page)
These servers are intended to host production and development web sites, web applications, and databases. At least one of these servers is reserved for running programs with large memory requirements (up to 16 GB). For other kinds of computationally intensive jobs, the clusters are a much better choice. More details are available on this page.
Local Windows servers (home page)
These servers are intended to host shared Windows applications and provide an alternative to running these programs at the desktop. Some programs, like SAS and Matlab, for which we don't own Linux licenses, are also hosted here with multi-user licenses. More details are available on this page.
Alabama Supercomputer Center (ASC) (home page)
There are a number of high performance computing resources available, we describe a few of these resources in more detail below and comment on their utility in support for statistical genetics research.
- SGI Altix supercomputer
- Up-to-date description of hardware
- A shared memory supercomputer with large memory queues, offering a solution to the class of large memory problem that the Beowulf clusters and UAB BG don't solve very well.
- In other words, if your problem uses a great deal of memory (think > 4 GB) and not many CPUs, then this resource will be of interest.
- Dense Memory Cluster (DMC)
For more information, see this page, which is maintained by local users of this resource.