• SETI@home: An Experiment in Public-Resource Computing
• Scaling e-Commerce Sites

The SETI@home paper is very similar to mine. It describes a development project like my own, so it lacks a formal hypothesis. The topic is the same, so some background information is common to the papers. The differences arise from the different project scope. SETI@home is an application to data analysis, so more technical details are included. SETI Visualizations has more information on how SETI has been marketed to the public.

• Purpose: Analyze SETI data using internet computing power
• Scope:
  • arrange for data collection at telescope
  • package and distribute data to computers
  • collect and analyze results
• Background:
  • Other SETI projects
  • overview of public-resource computing
• Development:
  • Data collection
  • Data distribution and hardware for same
  • Automated analysis
• Hypothesis: SETI@home far exceded expectations for involvement
• Algorithms:
  • garbage collector
  • SETI@home analysis loops
  • server processing of results
• Data: collected at Arecibo telescope
• Analysis: various Fourier transforms
• Conclusions: public-resource computing
  • high computing-to-data ratio
  • parallelism
  • error toleration
  • attractive
• References: 11

Scaling e-Commerce Sites is a more research-oriented project. The authors want to find what server architecture works best. Therefore the paper contains more analysis and evaluation. The introduction includes some information on previous research although the paper makes clear how that research was insufficient.

• Purpose: What server architectures support scaling of web sites
• Scope: No modifications to server/DB engine
• Background:
  • What are e-commerce sites
  • Scaling research for static sites
• Development:
  • server structure
  • engines used
• Hypothesis: some server architecture can scale with e-commerce site demand without continually changing software engine
• Algorithms
  • Shortest Queue First
  • Shortest Execution Length First
  • Query-Write Scheduling
• Data: collected by logging
• Analysis: TPC-W Benchmark
• Results: Asynchronous scheduling works best in shopping mode
• References: 19