Experiments Ashu Sabharwal David Wetherall Srinivasan Seshan Milind
Experiments • • Ashu Sabharwal David Wetherall Srinivasan Seshan Milind Buddhikot Dirk Grunwald Ratul Mahajan Hariharan Rahul Ashu Sabharwal • • • Jitu Padhye Dan Stancil Suman Banerjee Vishnu Navda Carl Gunter Yongguang Zhang Rice University
Making Experimentation a Science • Issue: How do you make experimental results verifiable and “repeatable” ? • Proposals – – Publish data sets, annotated as much as possible Publish code, especially if using a publicly available platform Network Emulation Make the whole process of presenting results close to how a “proof” is presented Ashu Sabharwal Rice University
Where is the Platform Which Helps me ? • Issue: I have to spend years making a platform – – Unpublishable It does not do what I want it to do “They” do not provide me the APIs By the time I get the platform, it is useless. Commodity hardware does more. • Proposals: Little consensus (FPGA/GPU/GPP) – – Continued $ important to build platforms (NSF/GENI) Important to build human capital to extend/use the platform Roadmap for RF capabilities My take: If you need it, you will have to spend the time/money Ashu Sabharwal Rice University
Role of Simulations • Issue: Revisit ? – Models – System level modeling accuracy unestablished/unclear – How do we proceed for the new modalities (cognitive/DSA) • Proposal: – – Not be afraid of ns-2/802. 11 fiasco Simulations are great for indicating coarse trends Closer to emulation by using real data A combination of analysis, simulations & experiments Ashu Sabharwal Rice University
Cool Idea of Day • Grass-root efforts can be quite powerful – SETI@home – UC Riverside, use motion sensors for early earthquake detection • Can we somehow use currently deployed hardware to collect spectrum usage statistics – A lot of Wi. Fi devices – Programmable cellular devices – Sparse deployments of GNUradio like setups Ashu Sabharwal Rice University
- Slides: 5