Ashu SabharwalRice University At-scale Programmable Wireless Testbeds Ashu Sabharwal Director, CMC Rice University.

Ashu Sabharwal Rice University

At-scale Programmable Wireless Testbeds

Ashu Sabharwal

Director, CMCRice University

Ashu Sabharwal Rice University

At-scale Programmable Wireless Testbeds

Ashu Sabharwal, Ed Knightly, Behnaam Aazhang and Joe Cavallaro

CMC

Ashu Sabharwal Rice University

Broadband Wireless Testbeds

• Programmable– Clean-slate design

• At-speed– Data-rate which users care about (10-100 Mbps)

• At-scale – 1000s of Real users

Ashu Sabharwal Rice University

Broadband Wireless Testbeds

• Programmable– Clean-slate design– High-performance

• At-speed– Data-rate which users care about (10-100 Mbps)– Stay with the technology curve

• At-scale – 1000s of Real users– Deployment: capacity-coverage-cost tradeoffs

WARP

TFA-Rice

Ashu Sabharwal Rice University

Wireless Open-Access Research Platform

• Xilinx Virtex 2Pro• 4 Daughtercard slots (upto

4x4 MIMO)• Fully operational & all

code open-source at warp.rice.edu

Ashu Sabharwal Rice University

Wireless Open-Access Research Platform

• Performance• Flexibility• Deployability

• Xilinx Virtex 2Pro• 4 Daughtercard slots (upto

4x4 MIMO)• Fully operational & all

code open-source at warp.rice.edu

Ashu Sabharwal Rice University

WARP: Performance

• Current PHY (2x2 MIMO)– 30 Mbps in 12.5 MHz– 60 Mbps tested in lab

• With optimizations, capable of hitting 100+ Mbps

Ashu Sabharwal Rice University

WARP: Flexibility

• Spectrum usage with 4 radios– 160 MHz across 2.4GHz and 5 GHz – Upto 4x4 40 MHz MIMO (802.11n, WiMax)

• Many extensions– TI DSP: can implement a full cell phone– Video card: for raw video input (joint source-channel

coding)

Ashu Sabharwal Rice University

WARP: Deployability

• Implements PHY, MAC, routing on every node– Can implement real-time clean-slate full stacks– All layers use conventional tools

• 8”x8” deployable like small access points

NetworkStack

Devices

Ashu Sabharwal Rice University

TFA-Rice

• Houston Pecan Park• Low income community

(56% <$25K/year)• Empower via technology

– Education (36% graduation rate)

– Health (high chronic disease rates)

Ashu Sabharwal Rice University

TFA-Rice

• Scalability• Observable• “Reality”• Applications

• Houston Pecan Park• Low income community

(56% <$25K/year)• Empower via technology

– Education (36% graduation rate)

– Health (high chronic disease rates)

Ashu Sabharwal Rice University

TFA-Rice: Scalability

• Two-tier Architecture– Access-tier : Clients (home & mobiles) to mesh nodes– Backhaul-tier : Mesh nodes wirelessly hop to gateway

• Dramatically reduced deployment costs– Limited fiber & inexpensive nodes

Ashu Sabharwal Rice University

TFA-Rice: Scalability (Pilot Program)

• Operational since 2004– 2000+ users @ 1 Mb/s over 3 Km2

• Growing to 4x4 scale - 4000 users over 4 Km2

• City inviting CFP to cover 620 Miles2 of Houston• Projected Growth: 1500 companies, $1 Billion market by 2010

QuickTime™ and aTIFF (Uncompressed) decompressor

are needed to see this picture.

Ashu Sabharwal Rice University

TFA-Rice: Observable

• Mesh nodes – Single-channel– Multiple antenna types

• Deployment research – Node spacing, path length to gateways– Link versus network-wide behavior

Ashu Sabharwal Rice University

TFA-Rice: Reality Bite 1

• Research has to treat users like end-customers– Even if you made no promises– Police/Commuters are using TFA

• Key issue – Useful only if real users use it– Real users will use it only if the network is useful

(reliable)– Running experiments, reboots, unreliabilities not

tolerated even for free services

Reliability and Accessibility are critical

Ashu Sabharwal Rice University

TFA-Rice: Reality Bite 2

• Want a large user base ?– Devices have to be reliable, cheap like 802.11– Research devices can only be in small numbers

• Does not preclude innovation– Control and backhaul protocols– Applications - healthcare

Standard Access Protocol Key to Scale

Ashu Sabharwal Rice University

TFA-Rice: Reality Bite 3

• CAPEX ~$20K/Km2

• OPEX– ISP– Customer support– Maintain research devices– Deploy new protocols– Work with community

• Earthlink, Tropos handle such scale

OPEX >> CAPEX

Ashu Sabharwal Rice University

TFA-Rice: Applications

• Rice, Methodist Hospital & TFA Collaboration• Addressing Chronic Cardiovascular Disease

– CVD responsible for 40% of all dealths in Texas– Higher occurrence in low-income

Ashu Sabharwal Rice University

TFA-Rice: Applications

• Rice, Methodist Hospital & TFA Collaboration• Addressing Chronic Cardiovascular Disease

– CVD responsible for 40% of all dealths in Texas– Higher occurrence in low-income

• Dasco hypothesis: 1950’s medical literature for low-cost monitoring – Simple metrics + info-gap theory yield effective predictors

Ashu Sabharwal Rice University

TFA-Rice: Applications

• Inter-connecting low-cost sensors, “health-phone,” pervasive wireless, and medical practitioners

• Collection, dissemination, and analysis of health data• Analysis alerts patient and community health worker of

potential problems

Health-phoneBlue Box

Ashu Sabharwal Rice University

Near Future Plans

• Deploy WARP nodes with novel new protocols• Higher throughput backhaul• Target 3 Mbps/user

Ashu Sabharwal Rice University

Conclusions

• Research challenges abound– Design, deploy, measure, model, …– Transformational applications

Ashu Sabharwal Rice University

Conclusions

• Research challenges abound– Design, deploy, measure, model, …– Transformational applications

• What is our ultimate metric ?– CANNOT just be bits/sec

• Improving – Life expectancy– Education

Ashu Sabharwal Rice University

More Information

• WARP : Ashu Sabharwal (warp.rice.edu)

• TFA-Rice : Ed Knightly (tfa.rice.edu)

• Health Phone: Lin Zhong (www.ruf.rice.edu/~mobile)

• Team: Behnaam Aazhang, Joe Cavallaro (cmc.rice.edu)