[ Up ] [ Current Status ] [ Previous Status ]

NT-Based Software Development and Training for Genome Analysis -Q4 1998 Status Report

February 15, 1999

The Intel investment of three servers, ten workstations, and a 100 Mbit network switch was matched by contributions from other sources:

The workstations and one server are used for classroom instruction and laboratories. Outside of class hours they are available to Plant Breeding staff and students for their own research. The other two servers are used for experimentation, especially related to porting the ACEDB genome database software from Unix to Windows NT.

This facility has made a big impact on the Plant Breeding Department, which outside of our own workgroup contains only two Windows computers. And it enabled the creation of Cornell's first courses in the growing area of Bioinformatics, drawing interest from all over campus.

Our success in the projects described below, and the advice of our collaborators in the Cornell Theory Center's Computational Genomics Institute, led us to choose Intel when we needed some powerful hardware to serve our databases to the Internet. Four quad-Xeon Dell PowerEdge 6300 servers are now installed at the Theory Center, and over the next months the Agricultural Genome Information System will be moved to them from its current site in Beltsville MD.

Research

Hugh Gauch is using one of the Intel computers for analyzing QTL (quantitative trait locus) experiments of several professors and graduate students. He is also developing software to handle QTL by environment interactions, to test QTLs for statistical significance, and to calculate least-squares estimators and superior predictors for the phenotypic trait data (such as yield) used to produce QTL scans. This work involves simulations and permutation tests with heavy computational demands, so the Pentium II workstation has been ideal.

We have done extensive alpha and beta testing of WinAce, the Windows 95/NT port of the ACEDB software. Each workstation has a copy of the executable and accesses the databases (GrainGenes, SolGenes, RiceGenes, RiceBlastDB, RoseDB, CabbagePatch) from a mapped drive on the server. Results from this configuration are very satisfactory, thanks in large part to the 100 Mbit network switch. However it is not an appropriate solution, since the entire database (some tens of megabytes each) must be moved through the network to local memory. Under Unix the equivalent configuration (databases on NFS-mounted disks) has been discarded as the worst-performing of several approaches to client/server ACEDB functionality. The better approaches are now being ported to NT and we will be testing them and putting them in service as they become available.

Instruction

This fall, two new graduate-level modules were developed and taught, making heavy use of the new NT machines. The lectures and schedule for these courses can be found at http://ascus.cit.cornell.edu/PB607/.

Each module was attended by approximately 20-25 people, with 15 students taking them for credit. Twice a week the instructor gave online demonstrations and lectures, and the third period was an open lab when students used the computer. The facility was open from 10 - 5 daily, and at any given time several students would be using the machines to complete their assignments. While very few of the students had any prior exposure to NT, no one had trouble transitioning to that platform.

The assignments primarily required students to access and interpret biologically relevant WWW resources, however, two database applications, GeneFlow and RiceGenes, were installed locally on the new machines for the students to use. The applications installed and performed without problem.

Almost all of the machines exhibited intermittent memory errors at first. This was solved by removing Internet Explorer.

Last modified on: 10/12/99