As a result, I spent several weeks in training session at NCSA during 1986, 1987 and 1988, and developed several afternoon seminars and pieces of documentation on NCSA hardware and its use. For example:

• Using the NCSA Crays from the University of Kansas
• Introduction to UNICOS

In 1990, KU decided to upgrade its central computing facility and I prepared the benchmark plan for the Request For Proposals (RFP). The winning system was a DEC VAX9000. I prepared writeups on the general use of this system, as well as the use of its vector-processing features.

In 1992, KU was loaned RISC systems by both DEC and IBM, and I became the system administrator of the first UNIX boxes at Academic Computing. I prepared several writeups describing their use

• Introduction to Ultrix
• Introduction to AIX

Later, I conducted a comparison of these systems, prepared a white paper describing the results, "A comparison of DEC, IBM and HP RISC systems," and presented a talk on general RISC performance, "RISC: The exciting new distributed computing game from IBM".

In early 1992, I recommended replacing our VAX9000 with a collection of RISC systems in "Delivering High-performance Computing in the Distributed Decade." and co-wrote bid specs for doing this in 1993.

In 1995 we once again began to investigate providing additional high performance computing capacity for the campus. Under the auspices of the Kansas Computer Planning Committee (KCPC), I developed and conducted surveys of both the presence and use of high performance computing within the State and prepared reports:

• Kansas high-performance computing questionnaire summary, and
• High-performance computing needs in Kansas: Availability, needs and sizing,

I developed the KCASC web site to assist users of the Origin and wrote documents to describe the new system in detail:

• The architecture and use of KU's Origin 2000
• Introduction to the Message Passing Interface (MPI) using Fortran

and later supervised the development of:

• Introduction to the Message Passing Interface (MPI) using C
• An Introduction to OpenGL

I also developed several scripts to monitor and manage the use of our hardware:

cpu-use.pl and cpu-and-error.pl

identify which users were assigned to which processors at any particular moment (see sample output),

make-daily.pl and make-monthly.pl

display daily and monthly accounting records, and

corral.pl

monitor job submission to enforce job submission policies.

In late 1998, I constructed a 5-node Beowulf system, called the "Jayohawk," to explore performance and operational issues related to Beowulf architectures, resulting in a CHECK conference presentation: "The Beowulf model for high-performance computing."

Parts of the Jayohawk were cannibalized to build a faux-Access Grid Node which preceeded the construction of our fully-featured Grid Node for use during the Kansas Edition of "Alliance Chautauqua 2000". The Chautauqua was one of a series of yearly conferences hosted by the National Computational Science Alliance (NCSA) using a combination of local and remote participants linked via the Alliance Access Grid video conferencing system.

I speced and acquired the components required to build our node here at ACS and served as Technical Director for the Chautauqua, assisted by Jeff Long and Taylor Kemper also of ACS.

In 2000 the Origin was upgraded to a 64-processor Origin 2400, with 16GB of main memory and 300GB of disk, and I served as system manager, interfacing between our systems group and the user community, for the duration of the grant.

Over the years I gave several talks on supercomputing to various groups as outreach efforts, including:

• Supercomputing resources on the Internet (Willets group)
• Supercomputing: Exploiting Concurrency for fun, profit, and science (KU chapter of ACM)