Bridging the Digital Divide: Technology for the “other” 4 Billion

1. FIT Bridging the Digital Divide: Technology for the “other” 4 Billion Umar Saifumar@mit.eduLUMS Computer Science Department

2. Digital Divide • Pakistan • Population: 160 Million • Sixth Most Populous in the world • Computer Users: < 9% • Internet Users: < 5%

3. Microsoft Digital Inclusion Program • First program specifically targeted at digital divide • Announced in Oct, 2005 • Awards announced on Jan, 2006 • Total Award: $1.2 Million • Our proposal was one of the 17 funded by the program • 162 Submission from 34 countries • 40 reviewers for selection • Recipients from 10 countries

4. Digital Inclusion Recipients • Guillermo Marshall, University of Buenos Aires, Argentina • Carlos Osvalod Rodriquez, Cequinor/Lanais EFO, CONICETUNLP, Argentina • Henry Nyongesa, University of Botswana, Botswana • Srinivasan Keshav, University of Waterloo, Canada • Miguel Nussbaum, Universidad Catolica de Chile, Chile • NyiMaTraShi, Tibet University, China･ • M.B. Srinivas, International Institute of Information Technology, India･ • Jan Carel Diehl, Delft University of Technology, Netherlands • Umar Saif, Lahore University of Management Sciences, Pakistan･ • Eduardo Grampin, Instituto de Computacion, Universidad de la Republica Uruguay, Uruguay･ • Thomas Anderson, University of Washington, U.S.･ • Suman Banerjee, University of Wisconsin-Madison, U.S.･ • John Bennett, University of Colorado at Boulder, U.S.･ • John Canny, University of California, Berkeley, U.S.･ • Joseph Rosen, MD, Thayer School of Engineering, Dartmouth College, U.S.･ • Roni Rosenfeld, Carnegie Mellon University, U.S.･ • Daniela Rus, Massachusetts Institute of Technology, U.S.

5. Motivation Developed World Developing World 2 MB Internet Connection < $40 2 MB Internet Connection > $4000 Bulk Data Transfer on the Internet > 70% Bulk Data Transfer on the Internet < 15% Digital Divide Average End-user Bandwidth via ISP > 100 kb/sec Average End-user Bandwidth via ISP < 10 kb/sec

6. Internet in Pakistan • Facts of life in the developing world • Internet used over dialup • ISP’s buy less, sell more • Economics • Politics • Lack of expertise • “Scratch card” provisioning

7. Internet in Pakistan • Average Dialup Bandwidth • Less than 10 kb/sec • Almost Never Used for • Exchanging • Disseminating • Accessing …. Content larger than a couple of hundred kibobyes • When was the last time you sent someone a 3.5 MB PDF file as an email extension?

8. How I Stumbled Upon this? • “Good research solves real problems in a practical way” • Started last year when I wanted to exchange a 3.5 MB PDF file with my dad • Two laptops sitting next to each other • No way to exchange data if you don’t have portable storage! • We actually went our and bought a CDR to exchange data….

9. Problem <10kb/sec Internet ~ 56kb/sec

10. Solution Bypass the Internet when exchanging large Internet ~ 56kb/sec

11. Email Attachments • Time to exchange a 3.5 MB file on the Internet ~ 1 hours (16 Kb/sec) • 30 mins upload and download • Assuming no disconnections • Time now (40 kb/sec) • 12 mins!!

12. Disruptive Technology • Of course Internet also started as an overlay over the phone lines • A new kind of Internet • Reminiscent of Pre-Internet days • FidoNet • UUCP

13. Why is this Practical? • Phone bills are becoming “Flat” • Rs 300/month -- free local calls • As long as you can identify a “close-by” host, “broadband access” is free • P2P systems already follow a similar model • MIT Chord, Microsoft Pasta

14. Dialup P2P-ISP Interleaving Key Idea: Use Internet as a directory service, not as digital pipe Internet ISP ISP Line-speed (~40kb/s) dialup connections Peer-to-peer dialup connections Dialup Underlay

16. Internet Client Applications Content-Push Applications Application Plugin Plugin Suspend-resume Session Layer DHT Caching, Indexing and Lookup Push subsystem Interleaving Heuristics Pre-fetching heuristics Operating System Bypass Protocol Stack Dialup Link Layer Our P2P Dialup Architecture

17. Three Evolving Applications • P2P file-sharing • Unrestricted Email exchange • Web-browsing …. in increasing order of difficulty

18. Making Bittorent Work Here • 30% of the Internet traffic • Less than 5% in Pakistan • Tit-for-tat Inherently “unfair” to dialup users • Make bittorent “sequential” • Sequential instead of parallel downloads • Backward compatibility by using a parallel “tracker” net • Our tracker does match-making of “offline” hosts

19. Dialup Bittorent

20. Email Access • Email attachment transfer will be layered on top of p2p access • Attach a file • “Securely” publish it on p2p • Recipient downloads in piece by piece

21. Web Access • Trickier • Must emulate connected behavior • Do a recursive pre-fetch: Tools already exist • When to interleave p2p-ISP? • Scheduling policies based on user browsing patterns • Suspend-resume Download Manager

22. Looking Ahead • Several follow-on projects • Several collaborative efforts within the Digital Inclusion recipients • We are trying to put together a proposal with Berkeley • Launching a multi-disciplinary effort at LUMS for developing world ICT

23. Rural Networking

24. Rural Networking

25. Multi-user devices

26. ImmunoSensors

27. ChoupalLink Inverse multiplexing over GSM/GPRS/ High-bandwidth Virtual Channel Figure 1: Inverse Multiplexing over cellular connections

28. ChoupalLink

29. ChoupaLink

30. Our Teleputer

31. Teleputer • Zero-configuration • Text-free Interface • Sensor-actuator • Cell-phone integrated • Shared Computing • Server-style processing

32. Teleputer Sensors

33. Teleputer Operation

34. Thank You!How Can you Help: Join http://www.dritte.orgQuestions: umar@mit.edu