Computing Resource Use Policy for SFSU Center for Computing for Life Sciences

M. Wong, D. Petkovic November 2006 (DRAFT)

Overview

The Center for Computing in Life Sciences (CCLS) http://cs.sfsu.edu/ccls/index.html provides focus and support for research in the area of computing for life sciences at San Francisco State University (SFSU). CCLS also provides access to high-performance computing resources—such as the Dell high performance cluster (HPC) computer—for participating faculty and students. This document describes the policies and procedures governing access to and use of those high-performance computing resources at CCLS. Such policies and procedures are necessary to ensure that the facilities are fairly shared, properly used, and effective in supporting the research and educational needs of their users. Policies outlined in this document may require occasional revision. Such changes are the responsibility of the CCLS IT committee.

About the CCLS Clusters

CCLS maintains two cluster systems, one for high performance computing, and another for instructional purposes. The HPC cluster is a Dell PowerEdge Blade Server system. More information on the Dell HPC cluster can be found at http://ccls.lab.sfsu.edu/bin/view/Cluster/DellPowerEdgeCluster.

The instructional cluster is comprised of donated hardware. The instructional cluster is ineffectual as a compute cluster, but very useful as a representative system for students to learn the concepts of parallel computing, and as a development system for the Dell HPC cluster. More information on the instructional cluster can be found at http://ccls.lab.sfsu.edu/bin/view/Cluster/InstructionalCluster.

Obtaining Access

All CCLS high performance resources are allocated on a per-project basis to a single principle investigator (PI) involved in an approved (not necessarily funded) CCLS project. Any faculty or staff member of the College of Science and Engineering at SFSU can serve as CCLS project PI. PIs can also request access for their participating students. All requests for allocations are to be made through the CCLS Allocations Request System.

There are two types of usage allocations: research and instructional.

Research project allocations support the computational needs of the PIs, their student researchers, and collaborators. Research project allocations begin when the project is approved as a CCLS project and last for the duration of the project. CCLS-approved research projects may be funded or unfunded and may include student culminating experience projects in the computing-for-life-sciences area.

Instructional project allocations support academic classes, which need high performance cluster computing. Instructional project allocations will last for the duration of the class plus sufficient time to prepare the class and summarize and grade projects.

Requests for allocations on all CCLS resources must be made using the CCLS Allocations Request System. This includes CCLS high performance computing systems and storage systems. If awarded an allocation, the PI will receive a login and initial password with which to access the appropriate resource. The PI can then direct other members of his/her research group to request access via CCLS Allocations Request System, according to approved usage polices and rules. The PI will also be awarded a standard allocation of 10GB of disk space for data, and default processing priority. PIs can request additional disk space and/or higher processing priorities. If the resources are available, CCLS shall do its best to honor these requests on a temporary basis. Each PI is responsible for ensuring that all users on the project allocation use CCLS resources in accordance with the CCLS Usage Policies.

PIs retain access allocation for as long as they are affiliated with CCLS, although for security reasons, unused accounts logins (by 6 months or more) may be temporarily disabled. To re-enable a disabled account login, simply contact CCLS staff. Students retain access allocations provided by CCLS or the PIs for as long as they are affiliated with the research.

Security

Users must maintain the security of CCLS systems and resources:

• login identifiers belong to a specific individual and must not be shared under any circumstances
• login identifier passwords must never be revealed to others
• PIs are responsible for notifying CCLS when login identifiers are created or deactivated

Suspicious login identifiers or processes may be terminated without notice, at the discretion of the CCLS staff.

Utilization and Load

CCLS staff has documented the purpose and proper use of high performance resources and services via a series of user guides and other web-based documents available on the CCLS collaborative web site: http://ccls.lab.sfsu.edu. Users assume the responsibility for becoming familiar with this documentation, particularly relating to issues such as adherence to resource limits, proper and considerate use of the batch queuing system, understanding the purpose and proper use of file systems and storage services, and proper use of CCLS facilities. These guidelines and provisions are in addition to the campus IT Resources Acceptable Use policy.

These CCLS allocations should be used only for work directly related to the project for which the allocation was requested and granted. CCLS systems, facilities and services should only be used for documented or obvious intended purpose. Examples of improper use follow:

• Using the high-performance systems to run applications that can be executed on basic PCs and workstations,
• Playing games or viewing movies not related to the research project,
• Using storage systems for backing up data that is not related to the project allocation.

In general, requests to install software released under liberal licenses, such as so-called Open-Source software, will be honored by the CCLS staff. Proprietary software must be approved by CCLS staff before installation on CCLS computing resources. This approval process includes, but is not limited to, (1) a feasibility review; (2) a computational impact analysis; and (3) an internal licensing audit. The CCLS computing resources shall be load balanced to afford equitable computing cycles to all users. Although the software for scheduling jobs on the CCLS computing resources has been configured to prevent a single user or a single project from dominating system resources, it may be possible to acquire a temporary increase in resource allocations. Requests for such special needs must be coordinated by the petitioner, and must be submitted to the CCLS IT staff. Such requests will be honored by the CCLS staff whenever possible. Last-minute requests to meet imminent deadlines are discouraged.

Owners of SW used in CCLS high performance cluster assume the responsibility of monitoring their application execution and using the high performance systems with tested and bug-free applications as much as it is possible so as to not interfere with the work of others.

Owners also assume the responsibility of using the SW with appropriate licenses. CCLS does not assume any responsibility arising from improper licensing issues of the SW. Processes that lock computing resources or otherwise prevent equitable distribution of computing cycles to all users will be identified and owners notified immediately for action. Such usage is strongly discouraged, and aforementioned processes may be terminated without notice, at the discretion of the CCLS staff.

User Support

CCLS users are encouraged to request assistance from CCLS staff when necessary. All requests for support must be submitted through the CCLS Consulting Request System. Special project collaborations are examples of situations requiring direct communication between CCLS staff and the user community. CCLS support staff will provide advice to users regarding the development, porting, debugging, and optimizing of codes, and in the proper use of CCLS resources. Support does not include the actual development, porting, debugging, and optimization of code. Support involvement shall be determined by the availability of CCLS staff. User support requests can be submitted through the CCLS support request webpage.

CCLS User News

Information regarding important CCLS activities, including system availability and upgrades, CCLS training classes, allocation renewal notifications, and holiday consulting coverage will be communicated to the user community via CCLS User News. Upon receiving an allocation the PI is automatically subscribed to all relevant CCLS User News categories. Users may unsubscribe from CCLS User News, but still assume the responsibility for the information contained therein. CCLS User News can be accessed through the CCLS collaborative web site: http://ccls.lab.sfsu.edu

Acknowledgement in Research Publications

Users are requested to acknowledge the use of the CCLS resources in publications of research, including (for the purposes of full disclosure) the vendor and cluster configuration. As per standard academic polices and practices, users should not make comparison and statements about various commercial systems, but can simply state how such systems have been used. For example: The experiments discussed in this paper were conducted using a 10-node Dell PowerEdge cluster, each with two 2.0 GHz dual-core processors and 4GB of RAM.

Enforcement

CCLS resources are deployed, configured, and operated to serve a large diverse user community of SFSU faculty and students and their collaborators engaged in CCLS projects. It is important that all users are aware of and abide by these CCLS use policies. Failure to do so may result in suspension or cancellation of the project and associated allocations and closure of all associated logins. Illegal transgressions will be addressed through SFSU and/or legal authorities.

Important links

• http://ccls.lab.sfsu.edu/cluster has links to the cluster documentation, policies, user support, the allocation request system, and support request system.

Contacts

Please direct comments and questions to Mike Wong <> or call at (415) 405-2119