CPSC 330 - Fall 2007 Project 1 - System Recommendation based on Benchmark Results Due at class time on September 19 You should work in teams of four. You should prepare an eight-to-ten-page recommendation report for a computer system for bioinformatics computing. Your recommended system(s) should be based on one of four assigned processor families: 1) Intel x86 (e.g., Core Duo, etc.) 2) AMD (e.g., Opteron) 3) IBM POWER or PowerPC 4) Intel Itanium You should also be prepared to give a brief overview of your recommendations in class (approx. 10-12 minute presentation per team). Your team's report should be a combination of unbiased observation and processor-family advocacy. Observation: * Include in your report an introductory section that gives an overview of bioinformatics and any major subfields within bioinformatics (which you can refer to when explaining the benchmarks you cite). Be sure to clearly explain workload characteristics of bioinformatics applications. (That is, how much floating-point, I/O, etc. - provide citations.) * You should include all relevant benchmarks (including ones in which your processor loses) in an appendix. Advocacy: * In your recommendations section, you can advocate why your processor family is better than the others. * In your recommendations section, you can highlight those benchmarks in which your processor wins and attempt to explain why those benchmarks are somehow better or more representative. (That is, make your best case, acting in this section as a marketing person.) (If your processor is a clear loser across most or all of multiple benchmarks, see me to discuss how to handle.) Assumptions: 1) You have a hardware budget of $150,000. 2) The system(s) you propose must be currently available and must use your assigned processor. 3) Although unrealistic, you do not have to include software and TCO costs in your project budget. 4) Your recommendation should logically derive from a careful analysis of benchmark results. You may mention scalability as a secondary factor, but you should not base your recommendation on perceived or hoped-for scalability. E.g., some old benchmark results (but don't base your recommendation on these numbers - use newer information) Portable Cray Bioinformatics Library - Long (ARSC), March 2004 cb_searchn() - gap-free nucleotide search allowing mismatches (http://recomb04.sdsc.edu/posters/jlongATarsc.edu_39.pdf) Xeon Athlon Power 4 Itanium 2 2.8 Ghz 1.4 GHz 1.7 GHz 1.3 Ghz cb_searchn / icc 1.0 0.71 2.3 1.18 HMMER - Purkayasta, et al. (TACC), May 2004 (no special optimizations) (http://www.tacc.utexas.edu/publications/performancehpcclusternodes.pdf) Xeon Opteron G5 Itanium 2 3.2 Ghz 2.0 GHz 2.0 GHz 1.5 Ghz serial 1.0 1.04 1.11 1.45 dual-threaded 1.0 1.3 0.96 1.56 Here are discussions and examples of recommendation reports (Please note that I want you to provide an "executive summary", rather than an "abstract" or just an introduction.) http://www.esm.vt.edu/design/assign/FinalReport.htm http://www.io.com/~hcexres/textbook/feas.html http://mistsy.home.texas.net/acc/tw/5.htm http://www.themfactor.com/articles/writing-executive-summaries-reports-recommendations.htm http://www.themfactor.com.au/Articles/writing-recommendations-proposals-reports.htm http://www.rrcc.edu/english/samplereport.html http://www.technical-writing-course.com/type-of-technical-report.html http://www.llnl.gov/tid/lof/documents/pdf/245207.pdf Feel free to "fire" a slacker who won't meet or participate by reporting them to me - I'll find some individual task for them to have to do to get a project 1 grade. (It will be a burdensome task, in order to provide an incentive for each group member to participate in the group.) It's best not to burden actual salespeople with emails and phone calls; their time is better spent working with paying customers. Specifically, in your preparation for the report, you cannot represent yourself as an employee of Clemson and/or inquiring on behalf of Clemson or any unit of Clemson University.