On the Advanced Cyberinfrastructure Teaching Facility (ACTF), learners and instructors explore computation at the big-data end of the spectrum, in topics ranging from Bioinformatics to Data Science.
The ACTF, hosted by the CQLS, provides high-performance-computing resources dedicated to instructional needs. These include:
- 300+ CPU cores across multiple high-core-count, high-memory machines
- 2+ TB of RAM
- 100+ TB of networked, high-speed ZFS file space
- 10GbE network throughout
- GPU-Compute resources
- Web-hosting resources
- Software management for instruction
- One-on-one support for instructors
To inquire about utilizing the ACTF for your course, or guidance on developing curricula to most effectively use these high-performance resources, please contact Brent Kronmiller.
Student Projects (utilizing the ACTF)
Grape Gene Coexpression Analysis: Martin Pearce, Aayushi Manchanda, and Yvonne Chang, mentored by Laurent Deluc
Using a systems biology approach, Martin, Aayushi, and Yvonne analyzed gene expression data from three tissues of grape berries (skin, pulp and seed) with Weighted Gene Co-Expression Network Analysis (WGCNA) to identify hub genes and modules associated with ripening processes.
Tree Stem Segmentation Using OctNet: Robert Schriver and Chu Qi
Utilizing GPU-enabled deep learning, this project aimed to automatically identify and measure tree trunks from images for forest management purposes.
ACTF Guidelines and Policies
Access and Usage
The ACTF is dedicated to instructional use for classes, workshops, and student clubs. Research or personal projects are strictly disallowed, unless they serve as a project for a class, workshop, or club.
Instructor setup and onboarding is handled by Brent Kronmiller; email Brent if you wish to use ACTF resources for your instructional needs. Once a space has been created, students may request access to it by filling out the form at https://shell.cqls.oregonstate.edu/access/actf.php.
User details (username, first name, last name) must be available to CQLS staff. Normally this happens as a matter of course during student registration. Instructors should also be cognizant of which students have registered to access their space.
We generally turn off access to spaces after the class/workshop has ended. Although we don't delete previous class data, we may archive it to tape and request that instructors minimize space usage at the end of each class (see Resource Sharing and Usage).
Software
The ACTF hosts a large number of globally-installed command-line software packages. To accommodate the needs of instructors, this software set is *not* a mirror of that available on the CQLS research cluster. Upgrades and changes to this set do occur term-to-term, though we strive to not make adjustments mid-term.
As a result, it is important that instructors check for software availability, ideally well before the term starts, each term. If you need software packages or libraries that aren't present, CQLS staff will try to install them if we receive at least 2 weeks’ notice before the start of the term or class (though see 'Support' below as some fees may be applicable). We cannot guarantee support for all software, as some software is incompatible with other currently-configured packages.
Each class space comes with a class-specific location for locally-installed software and scripts (/ACTF/Course/
Resource Usage and Sharing
While we do not restrict resource availability per-user or per-class (in terms of number of CPU cores or RAM used), please remember that the ACTF is a shared resource accessed by dozens to hundreds of users. If one class uses all available resources, there are none left for others! If you are unsure the best way to meet your instructional needs with the resources available, contact Brent for a free consultation.
We recommend the following as good general practices:
- The login node (shell.actf.oregonstate.edu) does not restrict job size, but computational jobs that will run for more than a few minutes or use more than a couple CPUs or gigabytes of RAM should be submitted via SGE_Batch or SGE_Array. Users may also interactively login to backend nodes with qrsh for heavier computation. Small-scale computation (editing or copying files, running short-lived scripts, etc.) is ok on the login node. Running top on the login node will show resources used and available.
- When submitting jobs or qrsh sessions, it is important to specify an appropriate number of slots (CPU cores in our configuration) so the scheduler can properly distribute resources. Don't submit a 20-thread blast job and only request 1 CPU! Similarly, don't request 20 CPUs and only use 1. The SGE_Avail tool will report the resources available at any given time in an easy-to-read fashion. When using qrsh, you can additionally check resource usage by running top, and be sure to log back out when finished to free those resources.
- When usage is heavy, or if you have many or particularly heavy jobs to submit, we recommend using no more than 50% of currently available resources, and always leave a couple of slots available on each machine so that others may qrsh in to check on the progress of their jobs with top. Again, SGE_Avail and top are useful tools for determining current cluster usage.
- We allocate space on a per-class (or workshop or club) basis, with the needs of each class in mind. Although we can allocate large amounts of space for temporary usage (e.g. 3T), we request that instructors bring space usage back down to under 100G at the end for archival purposes.