John C. Linford > Publications > Undergraduate

Powered by CMSimple


Undergraduate

  • John C. Linford, David Ferro and Brian Rague. Capturing Unused Resources to Solve Parallel Problems.

    Abstract: This was the first computing cluster constructed by the Department of Computer Science at Weber State University and it outsized a cluster purchased for the department of physics by 36 CPUs. The project was overseen by Dr. David Ferro and Dr. Brian Rague from the Massachusetts Institute of Technology and member of the High Capability Computing and Modeling Group at NASA's Jet Propulsion Laboratory. Hardware and software were supplied by Sun Microsystems and Altair Engineering. I used 36 SunBlade 100 workstations with SunPCI II co-processor cards and two SunUltra Enterprise Server 450s. Each SunBlade 100 workstation contained a 500Mhz 64-bit UltraSPARC II CPU, 256MB RAM, two 30GB hard disks, a 10/100 Ethernet NIC and the SunPCI II card. Each SunPCI II card carried a 733Mhz Intel Celeron CPU, 128MB RAM and a 10/100 Ethernet NIC. Each Ultra Enterprise Server contained four 480Mhz 64-bit UltraSPARC II CPUs, 4GB RAM, 512MB RAID-5 storage and five 10/100 Ethernet NICs. Communication was through switched 100Mb Ethernet. I grouped the workstations in clusters of four, spaced them over six 3'x10' tables, and rack-mounted the UltraEnterprise Servers in a standard upright enclosure. The cluster used only the UltraSPARC CPUs, but future work was expected to incorporate the Intel CPUs to expand the cluster to 80 CPUs.

    During this project, I also explored popular HPC software. Both Ultra Enterprise Servers ran Solaris 9; the SunBlade workstations used both Solaris 9 and OpenSolaris 10 during the duration of the project. I experimented with two cluster management packages: Altair PBS-Pro (PBSP) and Sun Grid Engine (SGE), and explored two MPI implementations: MPICH versions 1.1 and 2.0 and LAM/MPI. I managed the nodes via RSA-keyed SSH and used NFS to export a set of common libraries and files to the nodes. I provided Kerberos single sign-on and integrated with the school's existing single sign-on to make the cluster highly-accessible to all computer science students. This research was presented at the school undergraduate research conference and demonstrated that near-optimum speedup could be achieved for most applications using up to twenty CPUs. Applications executing on more than twenty CPUs experienced I/O and communication bottlenecks.
  • Michael Ferguson and John C. Linford. Creating the Campus Compact Web Page: A Service Learning Project. Best Project.

    Abstract: We engaged in multidisciplinary research to benefits the community by working with the Utah Campus Compact (UCC), a nonprofit organization which engages students in "service learning projects" to benefit the community. The UCC invests the student time and effort which is typically spent in purely academic projects into real-world projects which benefit the community. Our goal was to explore applications of emerging web technologies to non-profit organizational needs. Since the UCC matches existing resources with community needs, we developed a three-tier web application, using the recently-developed Microsoft .NET 1.1 framework, ASP.NET and C#, to automatically match available resources with community needs. The system also managed events, announcements and registration of organizations and individuals involved in service learning projects. The project took first place at the school undergraduate research conference and was presented at the Utah State Conference for Service Learning.
  • John C. Linford. Javascript von Neuman Machine.

    This was my first pass at something vaguely like research, and this code was eventually published in Connecting with Computer Science by Anderson, Ferro and Hilton. The code is a simple VonNeuman machine which understands pointers, implemented in Javascript. Download the code.