Technology for Developing Regions Eric Brewer Tier Group, UC Berkeley Madrid, Spain May 14, 2010 Thanks to  Sonesh Surana,...
The “Base of the Pyramid” <ul><li>3-4 billion people </li></ul><ul><ul><li>Equivalent purchasing power < $2/day </li></ul>...
Traditional Development <ul><li>Very top down </li></ul><ul><ul><li>International agencies fund big projects </li></ul></u...
Actual  Development: Cellphones <ul><li>3+B cellular users worldwide </li></ul><ul><ul><li>Vastly outpaces TV, PCs,  </li>...
Bottom-up Financing
Remittances <ul><li>Money sent back home to relatives </li></ul><ul><li>Remittance flows: </li></ul><ul><ul><li>Philippine...
Microfinance: Grameen Bank <ul><li>Owned entirely by the poor </li></ul><ul><ul><li>Began in one village in 1976 </li></ul...
Bottom-up Businesses
Grameen Telecom Cellular in rural Bangladesh <ul><li>‘ Village Phone’ </li></ul><ul><ul><li>Buy phone, rent to your neighb...
<ul><li>Medical transcription </li></ul><ul><li>Audio in, text out </li></ul><ul><li>Customers in US, India </li></ul><ul>...
Urban vs. Rural <ul><li>Value of the cellphone limited by coverage </li></ul><ul><ul><li>Even more true with the Cloud </l...
Where are we? Manila, Philippines Edmonton, Canada
Part 2: Rural Connectivity
Why WiFi?  <ul><li>Very low cost due to huge volume </li></ul><ul><li>Unlicensed spectrum (mostly!) </li></ul><ul><li>Incr...
NSDI 2007: WiLDNet Results
New World Record:  382 km (with Ermanno Pietrosemoli)
New World Record – 382 km Pico El Aguila, Venezuela Elev: 4200 meters
Rural Telemedicine <ul><li>Aravind Eye Hospitals </li></ul><ul><ul><li>Tamil Nadu, India </li></ul></ul><ul><ul><li>5 hosp...
Aravind Eye Hospital Network <ul><li>Achieve 4-5 Mb/s per link </li></ul><ul><li>Video-conferencing </li></ul><ul><li>E-m ...
Real Impact <ul><li>Over 160,000 patients so far </li></ul><ul><li>Centers are cash-flow positive </li></ul><ul><li>Over 3...
Part 3: Smart Phones
Cellphones as Computers <ul><li>Cellphones in all urban areas… </li></ul><ul><li>Many advantages: </li></ul><ul><ul><li>Sm...
SmartPhone Diagnostic Device <ul><li>Idea:  collection of very simple sensors that connect to the  audio  jack  </li></ul>...
CellScope: Cellphone Microscope <ul><li>Cellphone camera + big lens = microscope </li></ul><ul><li>Diagnosis: </li></ul><u...
Education: English <ul><li>English as a second language is very valuable </li></ul><ul><ul><li>Arguably defines “middle cl...
<ul><li>Substantial vocabulary gains: </li></ul><ul><ul><li>Mean score (pre-test): 3.7 out of 12 </li></ul></ul><ul><ul><l...
Summary <ul><li>Technology has a larger role to play </li></ul><ul><ul><li>WE  can decide is this is CS or not </li></ul><...
Backup
Part 3: Electric Power
Poor Quality Power <ul><li>Spikes and Swells: </li></ul><ul><li>Lost 50 power adapters </li></ul><ul><li>Burned 30 PoE por...
What about UPS? <ul><li>“ Affordable” UPS systems ($300) are of  standby-type </li></ul><ul><ul><li>Primary source is grid...
What’s in an iPhone (3GS) <ul><li>600 MHz ARM processor </li></ul><ul><ul><li>About 1200 MIPS for 300mW </li></ul></ul><ul...
Apple iPhone 3GS Touch screen: $35.25 Total: $178 $24 $2.25 $1.30 $8.50 $13 $1.25 $14.46 $3
Some Example Devices <ul><li>Health care:  one per health worker </li></ul><ul><ul><li>microscope, ultrasound, vital signs...
What’s in an iPhone?
Supercomputer <ul><li>“ Access is the Killer App” </li></ul><ul><ul><li>Access to the cloud >> any device </li></ul></ul><...
Replace the PC? <ul><li>Cellphone accessories: </li></ul><ul><ul><li>Big screen (~200M sold per year) </li></ul></ul><ul><...
Conclusion <ul><li>Smart phones = </li></ul><ul><ul><li>PC + sensors + connectivity + portability </li></ul></ul><ul><li>A...
Decentralized Development <ul><li>Decentralized financing exists and is large </li></ul><ul><ul><li>$13B/year in remittanc...
Schools in Batanes Prof. Manuel Ramos Jr., UP
Other Tier Technologies <ul><li>Delay Tolerant Networking (DTN) </li></ul><ul><li>Phone-based Applications </li></ul><ul><...
Education: multiple mice per PC <ul><li>Most PCs are used by 5-8 kids </li></ul><ul><li>Educational value? </li></ul><ul><...
Power Solutions <ul><li>New power controller: </li></ul><ul><ul><li>Provides clean power (no UPS required) </li></ul></ul>...
Open Spectrum Matters <ul><li>The issue is not just cost </li></ul><ul><ul><li>It is freedom to innovate </li></ul></ul><u...
Local expertise: Training and Migration Equipment Supply Installation Management Maintenance Migration at Aravind TIER, UC...
 
Education: English <ul><li>English as a second language is very valuable </li></ul><ul><ul><li>Arguably defines “middle cl...
<ul><li>Substantial vocabulary gains: </li></ul><ul><ul><li>Mean score (pre-test): 3.7 out of 12 </li></ul></ul><ul><ul><l...
Remote Management <ul><li>Really don’t want to visit nodes to fix them </li></ul><ul><ul><li>Even if you live in the same ...
Hardware Faults at Aravind, 2006 *Conservative Estimate Instances * Description Total Downtime 63 Router board not powered...
WiLD: Slotted WiFi MAC <ul><li>Move to TDMA with coarse slots </li></ul><ul><ul><li>Rate-based flow control (due to high R...
Africa
International Bandwidth <ul><li>$3000 by Mb/s per month (in Africa) </li></ul><ul><ul><li>Almost all via satellite </li></...
WiFi Internet for Africa? <ul><li>Key idea: </li></ul><ul><ul><li>Put in a WiFi backbone </li></ul></ul><ul><ul><li>Conver...
SW Faults Software Faults at Aravind (in 2006) *Conservative Estimate Instances * Description Total Downtime 4 No default ...
2. Data Collection and Monitoring <ul><li>Designed Push-based PhoneHome system </li></ul><ul><ul><li>Measures node, link, ...
3. Alternate Network Entry Points <ul><li>Getting into the network is itself a challenge </li></ul><ul><ul><li>And once in...
4. Recovery Mechanisms <ul><li>HW watchdog </li></ul><ul><ul><li>Low voltage is still problematic for on-board HW watchdog...
Operational Results
Lesson 1: Reduce Need For Trained Staff <ul><li>Reduce downtimes through better power </li></ul><ul><li>Nodes reboot relia...
Lesson 2: Simple Re-design Is Often Enough <ul><li>WiLDNet MAC </li></ul><ul><li>Improved Low Voltage Disconnect </li></ul...
Lesson 3: Real Cost of Power Is Cleaning <ul><li>Grid power is bad quality </li></ul><ul><li>Cleaning it up is costly and ...
Thank you! <ul><li>Acknowledgments: </li></ul><ul><li>Ermanno Pietrosemoli, Venezuela </li></ul><ul><li>Alan Mainwaring, I...
Is it sustainable?   <ul><li>Vision center setup cost: $4000 (once) </li></ul><ul><li>Cost of Link: $1000 (once), towers $...
The Wireless Hypothesis <ul><li>Wireless infrastructure is the first viable infrastructure for rural areas </li></ul><ul><...
Schools in Batanes Prof. Manuel Ramos Jr. UP
SmartPhone Diagnostic Device <ul><li>Idea:  collection of very simple sensors that connect to the audio jack of a “smart p...
Upcoming SlideShare
Loading in …5
×

Brewer: Contributions of Social Research on ICT4D

898 views

Published on

E. Brewer, TIER Group: Contribuciones de la Investigación Tecnológica en las TIC para el Desarrollo Humano, Contributions of Social Research on ICT4D

Published in: Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
898
On SlideShare
0
From Embeds
0
Number of Embeds
125
Actions
Shares
0
Downloads
0
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Brewer: Contributions of Social Research on ICT4D

  1. 1. Technology for Developing Regions Eric Brewer Tier Group, UC Berkeley Madrid, Spain May 14, 2010 Thanks to Sonesh Surana, Kurtis Heimerl and the TIER Students & Faculty
  2. 2. The “Base of the Pyramid” <ul><li>3-4 billion people </li></ul><ul><ul><li>Equivalent purchasing power < $2/day </li></ul></ul><ul><li>Could swell to 6-8 billion over next 25 years </li></ul><ul><li>Most live in rural villages or urban slums </li></ul><ul><ul><li>Movement towards urbanization </li></ul></ul><ul><li>Education levels are low or non-existent </li></ul><ul><ul><li>Especially for women </li></ul></ul><ul><ul><ul><li>=> high birth rates => more poverty </li></ul></ul></ul><ul><li>Markets are hard to reach, disorganized </li></ul>
  3. 3. Traditional Development <ul><li>Very top down </li></ul><ul><ul><li>International agencies fund big projects </li></ul></ul><ul><ul><li>Often with stings and debt attached </li></ul></ul><ul><ul><li>Difficult to manage (e.g. corruption) </li></ul></ul><ul><li>Little role for high technology </li></ul><ul><ul><li>(some role for medicine, agricultural tech) </li></ul></ul><ul><ul><li>Agencies have almost no technical capacity </li></ul></ul><ul><ul><li>Technology chosen/used by large contractors </li></ul></ul>
  4. 4. Actual Development: Cellphones <ul><li>3+B cellular users worldwide </li></ul><ul><ul><li>Vastly outpaces TV, PCs, </li></ul></ul><ul><ul><ul><li>1.13 billion phones sold in 2009 (0.3 billion PCs) </li></ul></ul></ul><ul><li>+10% mobile penetration => GDP up 0.8% </li></ul><ul><ul><li>Causal, not just correlation </li></ul></ul><ul><li>Driven by bottom-up demand : </li></ul><ul><ul><li>Ease of use (voice) </li></ul></ul><ul><ul><li>Need for communications: work, remittances </li></ul></ul><ul><ul><li>Prepaid minutes and scratch-off cards </li></ul></ul>
  5. 5. Bottom-up Financing
  6. 6. Remittances <ul><li>Money sent back home to relatives </li></ul><ul><li>Remittance flows: </li></ul><ul><ul><li>Philippines $14B (13% GDP) </li></ul></ul><ul><ul><li>North Africa: $17.6B </li></ul></ul><ul><ul><li>West Africa: $10.4B </li></ul></ul><ul><ul><li>Central Africa: $2.7B </li></ul></ul><ul><ul><li>East Africa: $5.9B </li></ul></ul><ul><ul><li>South Africa: $2.0B </li></ul></ul><ul><li>Many small payments </li></ul><ul><li>Critical part of the economy </li></ul><ul><ul><li>Largely informal </li></ul></ul><ul><ul><li>Not always legal </li></ul></ul>$40B
  7. 7. Microfinance: Grameen Bank <ul><li>Owned entirely by the poor </li></ul><ul><ul><li>Began in one village in 1976 </li></ul></ul><ul><ul><li>2.6 million borrowers (95% women), over 1,000 branches in over 42,000 villages. 12,000 staff. </li></ul></ul><ul><li>Has loaned more than US$3.9B since inception </li></ul><ul><ul><li>Over US$3.5B repaid with interest (98.75% recovery rate); $290M loaned in the last 12 months. </li></ul></ul><ul><li>Has never accepted any charity </li></ul><ul><li>— has always been a profitable social enterprise </li></ul><ul><li>46.5% of borrowers have crossed the poverty line </li></ul>
  8. 8. Bottom-up Businesses
  9. 9. Grameen Telecom Cellular in rural Bangladesh <ul><li>‘ Village Phone’ </li></ul><ul><ul><li>Buy phone, rent to your neighbors </li></ul></ul><ul><li>So far over 95,000 loans of average US$200 have been given to buy mobile phones. </li></ul><ul><li>Covers 50,000 of 68,000 villages </li></ul><ul><ul><li>60M users </li></ul></ul><ul><ul><li>=> Scales! </li></ul></ul><ul><li>Phone Lady income up by 2x </li></ul><ul><ul><li>Maintains the system </li></ul></ul><ul><ul><li>=> Works! </li></ul></ul><ul><ul><li>Example of new rural income </li></ul></ul>
  10. 10. <ul><li>Medical transcription </li></ul><ul><li>Audio in, text out </li></ul><ul><li>Customers in US, India </li></ul><ul><li>Rural Kerela: </li></ul><ul><ul><li>30km from a small city </li></ul></ul><ul><ul><li>cheaper than Bangalore, Chennai, etc. </li></ul></ul><ul><li>Pays 5-8 cents/line </li></ul><ul><ul><li>about $8/hour </li></ul></ul><ul><li>Transcribe 3 times then merge! </li></ul><ul><li>WiFi enabled rural income </li></ul>
  11. 11. Urban vs. Rural <ul><li>Value of the cellphone limited by coverage </li></ul><ul><ul><li>Even more true with the Cloud </li></ul></ul><ul><li>Cellular is an urban phenomenon </li></ul><ul><ul><li>Rural areas generally left out </li></ul></ul><ul><ul><li>… basestation costs limit rural deployment </li></ul></ul><ul><li>This is the real “digital divide” </li></ul><ul><ul><li>Not between nations </li></ul></ul><ul><ul><li>… but between urban and rural (even in the US) </li></ul></ul>
  12. 12. Where are we? Manila, Philippines Edmonton, Canada
  13. 13. Part 2: Rural Connectivity
  14. 14. Why WiFi? <ul><li>Very low cost due to huge volume </li></ul><ul><li>Unlicensed spectrum (mostly!) </li></ul><ul><li>Incremental deployment </li></ul><ul><ul><li>Limited capital </li></ul></ul><ul><ul><li>Start small, grow over time </li></ul></ul><ul><li>Our innovations: </li></ul><ul><ul><li>New software for very long distances </li></ul></ul><ul><ul><li>Better use of spectrum </li></ul></ul>
  15. 15.
  16. 16. NSDI 2007: WiLDNet Results
  17. 17. New World Record: 382 km (with Ermanno Pietrosemoli)
  18. 18. New World Record – 382 km Pico El Aguila, Venezuela Elev: 4200 meters
  19. 19. Rural Telemedicine <ul><li>Aravind Eye Hospitals </li></ul><ul><ul><li>Tamil Nadu, India </li></ul></ul><ul><ul><li>5 hospitals </li></ul></ul><ul><ul><li>But too far for most to walk </li></ul></ul><ul><li>Need: </li></ul><ul><ul><li>15M blind in India </li></ul></ul><ul><ul><li>70% of blindness treatable </li></ul></ul><ul><ul><li>7% in rural areas get care </li></ul></ul><ul><li>Goals: </li></ul><ul><ul><li>50 rural vision centers </li></ul></ul><ul><ul><li>Diagnosis and prevention </li></ul></ul>
  20. 20.
  21. 21. Aravind Eye Hospital Network <ul><li>Achieve 4-5 Mb/s per link </li></ul><ul><li>Video-conferencing </li></ul><ul><li>E-m ail, training </li></ul><ul><li>6000 consultations/month </li></ul>Routers used : PC Engines Wrap boards, 266 Mhz CPU, 512 MB Cost : $140
  22. 22. Real Impact <ul><li>Over 160,000 patients so far </li></ul><ul><li>Centers are cash-flow positive </li></ul><ul><li>Over 30,000 patients have recovered sight </li></ul><ul><li>Growing to 50 centers covering 2.5M people </li></ul><ul><li>Hoping to replicate in other cities </li></ul>
  23. 23. Part 3: Smart Phones
  24. 24. Cellphones as Computers <ul><li>Cellphones in all urban areas… </li></ul><ul><li>Many advantages: </li></ul><ul><ul><li>Small, portable, with self-contained power </li></ul></ul><ul><ul><li>Voice works for all languages (but not text) </li></ul></ul><ul><ul><li>Easy to use, culturally accepted </li></ul></ul><ul><li>174M smart phones sold in 2009 </li></ul><ul><ul><li>15% of all phones (up from 12.8% in 2008) </li></ul></ul><ul><ul><li>Smart phones will pass PCs in the near future </li></ul></ul><ul><li>PCs may not ever get great penetration </li></ul>
  25. 25. SmartPhone Diagnostic Device <ul><li>Idea: collection of very simple sensors that connect to the audio jack </li></ul><ul><ul><li>Easy: heart rate, respiratory rate, temperature </li></ul></ul><ul><ul><li>Medium: blood oxygen, ECG, fetal heart rate </li></ul></ul><ul><ul><li>Hard: blood pressure </li></ul></ul><ul><li>Goal: much cheaper, easy to use diagnostic device </li></ul><ul><ul><li>Phone converts raw data into readings </li></ul></ul><ul><ul><li>Can also forward raw data over GSM </li></ul></ul>
  26. 26. CellScope: Cellphone Microscope <ul><li>Cellphone camera + big lens = microscope </li></ul><ul><li>Diagnosis: </li></ul><ul><ul><li>Image recognition on the phone? </li></ul></ul><ul><ul><li>… in the cloud? </li></ul></ul><ul><ul><li>… by a remote expert? </li></ul></ul><ul><li>Clinical quality: </li></ul><ul><ul><li>Malaria, </li></ul></ul><ul><ul><li>TB already verified as detectable </li></ul></ul>Dan Fletcher UC Berkeley
  27. 27. Education: English <ul><li>English as a second language is very valuable </li></ul><ul><ul><li>Arguably defines “middle class” in India </li></ul></ul><ul><li>Goal: teach English via cell phone games </li></ul><ul><ul><li>Cell phones are a better platform than PCs </li></ul></ul><ul><ul><li>Games can be effective with limited teachers </li></ul></ul><ul><ul><li>Must localize the content </li></ul></ul><ul><ul><li>Also developing continuous speech recognition on the phone </li></ul></ul>
  28. 28. <ul><li>Substantial vocabulary gains: </li></ul><ul><ul><li>Mean score (pre-test): 3.7 out of 12 </li></ul></ul><ul><ul><li>Mean score (post-test) : 8 out of 12 </li></ul></ul><ul><ul><li>p-value < 0.001 </li></ul></ul><ul><ul><li>std. dev. = 3.7 </li></ul></ul>Learning Assessment (Preliminary) Matthew Kam Divya Ramachandran John Canny
  29. 29. Summary <ul><li>Technology has a larger role to play </li></ul><ul><ul><li>WE can decide is this is CS or not </li></ul></ul><ul><li>Many needs: </li></ul><ul><ul><li>Connectivity, power solutions, applications </li></ul></ul><ul><ul><li>Capacity building: continuous training/support </li></ul></ul><ul><ul><li>A large role for phones </li></ul></ul><ul><li>Decentralized development works </li></ul><ul><ul><li>Driven by bottom-up funding </li></ul></ul><ul><ul><li>Inherently local solutions and buy in </li></ul></ul>
  30. 30. Backup
  31. 31. Part 3: Electric Power
  32. 32. Poor Quality Power <ul><li>Spikes and Swells: </li></ul><ul><li>Lost 50 power adapters </li></ul><ul><li>Burned 30 PoE ports </li></ul><ul><li>Low Voltages: </li></ul><ul><li>Incomplete boots </li></ul><ul><li>HW watchdog fails </li></ul><ul><li>Frequent Fluctuations: </li></ul><ul><li>CF corruptions </li></ul><ul><li>Battery Damage </li></ul>Voltage Range Number of Instances seen over 6 weeks
  33. 33. What about UPS? <ul><li>“ Affordable” UPS systems ($300) are of standby-type </li></ul><ul><ul><li>Primary source is grid </li></ul></ul><ul><ul><li>Secondary source is battery </li></ul></ul><ul><ul><li>Good quality power supplied only during outage! </li></ul></ul><ul><ul><li>2006 fault episodes include use of UPS </li></ul></ul>
  34. 34. What’s in an iPhone (3GS) <ul><li>600 MHz ARM processor </li></ul><ul><ul><li>About 1200 MIPS for 300mW </li></ul></ul><ul><ul><li>256MB ram, 16 GB flash </li></ul></ul><ul><ul><li>Roughly a 1999 Pentium III ~ 1350 MIPS for 37W </li></ul></ul><ul><li>PowerVR graphics core </li></ul><ul><ul><li>7M triangles/sec, 250M pixels/sec </li></ul></ul><ul><li>Mac OS X </li></ul><ul><ul><li>same as Mac, iPad, iPod Touch </li></ul></ul><ul><li>Sensors: </li></ul><ul><ul><li>Camera, mic, GPS, compass, proximity, ambient light, ambient noise, 3D accelerometer, touchscreen, temperature </li></ul></ul><ul><li>WiFi, GSM, Bluetooth, USB </li></ul>
  35. 35. Apple iPhone 3GS Touch screen: $35.25 Total: $178 $24 $2.25 $1.30 $8.50 $13 $1.25 $14.46 $3
  36. 36. Some Example Devices <ul><li>Health care: one per health worker </li></ul><ul><ul><li>microscope, ultrasound, vital signs, simple lab tests </li></ul></ul><ul><li>Education: </li></ul><ul><ul><li>Simple media players (one per teacher) </li></ul></ul><ul><ul><ul><li>Educational video </li></ul></ul></ul><ul><ul><ul><li>Wikipedia fits on 2 DVDs </li></ul></ul></ul><ul><ul><li>USB keys with TV out and buttons </li></ul></ul><ul><ul><ul><li>Read books on TV, take tests </li></ul></ul></ul><ul><ul><ul><li>One per student </li></ul></ul></ul><ul><li>Commerce: </li></ul><ul><ul><li>Simple point of sale devices (with a printer) </li></ul></ul><ul><ul><li>Field testing for crops </li></ul></ul>
  37. 37. What’s in an iPhone?
  38. 38. Supercomputer <ul><li>“ Access is the Killer App” </li></ul><ul><ul><li>Access to the cloud >> any device </li></ul></ul><ul><li>Cellphones are pocket supercomputers </li></ul><ul><ul><li>Every search </li></ul></ul><ul><ul><li>Image recognition, object recognition: </li></ul></ul><ul><ul><ul><li>Red laser </li></ul></ul></ul><ul><ul><ul><li>Google Goggles </li></ul></ul></ul><ul><ul><li>Sound recognition: dictation, translation… Shazam </li></ul></ul>
  39. 39. Replace the PC? <ul><li>Cellphone accessories: </li></ul><ul><ul><li>Big screen (~200M sold per year) </li></ul></ul><ul><ul><ul><li>… with wireless access (Wireless USB coming) </li></ul></ul></ul><ul><ul><li>Keyboard </li></ul></ul><ul><ul><ul><li>Or just use voice + supercomputer? </li></ul></ul></ul><ul><ul><li>Extra processing and disk space? </li></ul></ul><ul><ul><ul><li>Maybe…. Or just use the cloud </li></ul></ul></ul><ul><ul><ul><li>Likely want local graphics acceleration </li></ul></ul></ul><ul><ul><ul><li>PC as a cellphone accessory </li></ul></ul></ul><ul><li>My bet: cellphone wins… maybe it already has </li></ul>
  40. 40. Conclusion <ul><li>Smart phones = </li></ul><ul><ul><li>PC + sensors + connectivity + portability </li></ul></ul><ul><li>Access is the killer app </li></ul><ul><ul><li>The real win is phone+Cloud </li></ul></ul><ul><li>Cellphones are transforming development </li></ul><ul><ul><li>Driving increased quality of life for the poor </li></ul></ul><ul><ul><li>The first high-tech solution to cross over </li></ul></ul><ul><li>Rural access must be a priority </li></ul>
  41. 41. Decentralized Development <ul><li>Decentralized financing exists and is large </li></ul><ul><ul><li>$13B/year in remittances to Philippines </li></ul></ul><ul><li>Cellular and the Internet are enablers </li></ul><ul><li>Bottom-up projects WORK </li></ul><ul><ul><li>Tend to fit the actual needs </li></ul></ul><ul><ul><li>Inherently localized </li></ul></ul><ul><li>Policy should promote local development </li></ul><ul><ul><li>Open spectrum </li></ul></ul><ul><ul><li>Low taxation on IT equipment </li></ul></ul><ul><ul><li>Matching funds? Contests? </li></ul></ul>
  42. 42. Schools in Batanes Prof. Manuel Ramos Jr., UP
  43. 43. Other Tier Technologies <ul><li>Delay Tolerant Networking (DTN) </li></ul><ul><li>Phone-based Applications </li></ul><ul><ul><li>Health records & diagnostics </li></ul></ul><ul><ul><li>Voice-messaging phones </li></ul></ul><ul><li>Education </li></ul><ul><ul><li>Learning English from on a cellphone </li></ul></ul><ul><ul><li>Multiple Mice per PC improves learning </li></ul></ul>
  44. 44. Education: multiple mice per PC <ul><li>Most PCs are used by 5-8 kids </li></ul><ul><li>Educational value? </li></ul><ul><ul><li>Appears to reduce attrition </li></ul></ul><ul><li>Give each child their own mouse (joint work with MSR India) </li></ul><ul><ul><li>Significant impact on boys </li></ul></ul><ul><ul><li>… but not for competitive games </li></ul></ul><ul><ul><li>Most effective game: all users must pick correct answer </li></ul></ul><ul><li>New version works with legacy software! </li></ul>
  45. 45. Power Solutions <ul><li>New power controller: </li></ul><ul><ul><li>Provides clean power (no UPS required) </li></ul></ul><ul><ul><li>Supports solar or grid power </li></ul></ul><ul><ul><li>Disconnects power when voltage is too low </li></ul></ul><ul><ul><li>Smart charger for batteries: doubles battery life </li></ul></ul><ul><ul><li>Solar PPT (peak power tracking) => 15% more power draw </li></ul></ul><ul><ul><li>Simplifies remote management </li></ul></ul><ul><ul><li>Perhaps $50 </li></ul></ul><ul><li>Have not lost any routers in last year </li></ul><ul><li>Must plan around power issues </li></ul><ul><ul><li>IT needs high-quality power </li></ul></ul><ul><ul><li>Cheaper to buy power than to clean it ! </li></ul></ul>
  46. 46. Open Spectrum Matters <ul><li>The issue is not just cost </li></ul><ul><ul><li>It is freedom to innovate </li></ul></ul><ul><li>Rural areas can’t wait for carriers </li></ul><ul><ul><li>Freedom to innovate, deploy is critical </li></ul></ul><ul><ul><li>(No one buys rural spectrum intentionally!) </li></ul></ul><ul><ul><li>Interference is not a problem in rural areas </li></ul></ul><ul><li>WiFi is the right place to focus for open spectrum </li></ul><ul><ul><li>The lowest cost data solution </li></ul></ul><ul><ul><li>Lower training, deployment costs </li></ul></ul><ul><ul><li>Philippines should open up 5.8GHz </li></ul></ul>
  47. 47. Local expertise: Training and Migration Equipment Supply Installation Management Maintenance Migration at Aravind TIER, UC Berkeley Aravind Local Vendor Jan’06 – Jun’06 2007: 5 more clinic links Jun’07 – Dec’07 Jul’06 – Dec’06 Jan’07 – Jun’07
  48. 49. Education: English <ul><li>English as a second language is very valuable </li></ul><ul><ul><li>Arguably defines “middle class” in India </li></ul></ul><ul><li>Goal: teach English via cell phone games </li></ul><ul><ul><li>Cell phones are a better platform than PCs </li></ul></ul><ul><ul><li>Games can be effective with limited teachers </li></ul></ul><ul><ul><li>Must localize the content </li></ul></ul><ul><ul><li>Also developing continuous speech recognition on the phone </li></ul></ul>
  49. 50. <ul><li>Substantial vocabulary gains: </li></ul><ul><ul><li>Mean score (pre-test): 3.7 out of 12 </li></ul></ul><ul><ul><li>Mean score (post-test) : 8 out of 12 </li></ul></ul><ul><ul><li>p-value < 0.001 </li></ul></ul><ul><ul><li>std. dev. = 3.7 </li></ul></ul>Learning Assessment (Preliminary) Matthew Kam Divya Ramachandran John Canny
  50. 51. Remote Management <ul><li>Really don’t want to visit nodes to fix them </li></ul><ul><ul><li>Even if you live in the same country! </li></ul></ul><ul><ul><li>Remotes locations, risky travel </li></ul></ul><ul><li>Solutions: </li></ul><ul><ul><li>Robust engineering (e.g. watchdogs) </li></ul></ul><ul><ul><li>Need preventative maintenance </li></ul></ul><ul><ul><li>Remote diagnosis helps some </li></ul></ul><ul><ul><li>Alternate paths to remote nodes </li></ul></ul><ul><ul><ul><li>Sometimes we use cellular </li></ul></ul></ul>
  51. 52. Hardware Faults at Aravind, 2006 *Conservative Estimate Instances * Description Total Downtime 63 Router board not powered 63 days 7 Router powered but hung 10 days 21 Router powered but not connected to remote LAN (burned ethernet ports) 34 days 3 Router on, but wireless cards not transmitting (low voltage) 2 days 3 Router on, but pigtails not connected 45 days 1 Router on, but antenna Line-of-Sight blocked 8 weeks >90% of faults are power-related
  52. 53. WiLD: Slotted WiFi MAC <ul><li>Move to TDMA with coarse slots </li></ul><ul><ul><li>Rate-based flow control (due to high RTT) </li></ul></ul><ul><ul><li>Exponential backoff is bad for voice </li></ul></ul><ul><li>Move to bulk acks </li></ul><ul><ul><li>Turn off automatic acks </li></ul></ul><ul><ul><li>Normal ack timeouts too short </li></ul></ul><ul><li>Add some forward-error correction </li></ul><ul><li>Support multiple antennas per pole </li></ul><ul><ul><li>“ burst synchronization” => all send or all receive </li></ul></ul><ul><ul><li>Turn off carrier sense (to co-transmit with neighbor) </li></ul></ul><ul><li>Working on: </li></ul><ul><ul><li>better MAC for mixed voice/data </li></ul></ul><ul><ul><li>point-to-multipiont </li></ul></ul>
  53. 54. Africa
  54. 55. International Bandwidth <ul><li>$3000 by Mb/s per month (in Africa) </li></ul><ul><ul><li>Almost all via satellite </li></ul></ul><ul><li>What about fiber? </li></ul><ul><ul><li>Undersea fiber along the western side (SAT3) </li></ul></ul><ul><ul><li>But priced just below satellite!! </li></ul></ul><ul><ul><ul><li>Benin: cheap at $2000 for Mb/s/month </li></ul></ul></ul><ul><ul><ul><li>Nitel (Nigeria) was $8000 until recently </li></ul></ul></ul><ul><ul><li>Very big place to lay fiber… </li></ul></ul><ul><li>Need to decide intranet or Internet? </li></ul>
  55. 56. WiFi Internet for Africa? <ul><li>Key idea: </li></ul><ul><ul><li>Put in a WiFi backbone </li></ul></ul><ul><ul><li>Convert segments to fiber over time </li></ul></ul><ul><li>Much lower capital up front than fiber! </li></ul><ul><li>When link is full, you can fund fiber! </li></ul><ul><li>Crazy? maybe </li></ul>
  56. 57.
  57. 58.
  58. 59.
  59. 60. SW Faults Software Faults at Aravind (in 2006) *Conservative Estimate Instances * Description Total Downtime 4 No default gateway specified 4 days 3 Wrong ESSID, channel, mode 3 days 2 Wrong IP address 2 days 2 Misconfigured routing 3 days
  60. 61. 2. Data Collection and Monitoring <ul><li>Designed Push-based PhoneHome system </li></ul><ul><ul><li>Measures node, link, network level properties </li></ul></ul><ul><ul><li>Posts to server every 3 hours </li></ul></ul><ul><li>Also used for remote management </li></ul><ul><ul><li>With every post, opens reverse ssh tunnel </li></ul></ul><ul><li>Has helped with some diagnosis </li></ul><ul><ul><li>E.g. 2 wireless cards do not work together during low voltage situations </li></ul></ul>
  61. 62.
  62. 63. 3. Alternate Network Entry Points <ul><li>Getting into the network is itself a challenge </li></ul><ul><ul><li>And once in, other points may not be reachable </li></ul></ul><ul><li>Enable alternate points using Satellite (V-SAT) and Cellphone </li></ul><ul><ul><li>V-SAT up only 60% of time at Aravind </li></ul></ul><ul><ul><li>GPRS phone (console access) at AirJaldi </li></ul></ul><ul><li>Use in combination with cascaded hop-by-hop logins </li></ul><ul><ul><li>Advantage: Does not depend on system-wide routing or underlying IP configuration </li></ul></ul><ul><li>Can recover from SW configuration errors </li></ul><ul><li>No recourse for physical faults, except visits </li></ul><ul><ul><li>But might still gain useful information to help preparation of visit </li></ul></ul>
  63. 64. 4. Recovery Mechanisms <ul><li>HW watchdog </li></ul><ul><ul><li>Low voltage is still problematic for on-board HW watchdog </li></ul></ul><ul><ul><li>Use LVD, so the router is not hung, but disconnected </li></ul></ul><ul><ul><li>At Airjaldi, built external HW watchdog for < $1 for routers without on-board HW watchdogs </li></ul></ul><ul><li>SW watchdog </li></ul><ul><ul><li>Run by cron every 4 minutes </li></ul></ul><ul><ul><li>Configuration file lists what parameters to measure and what services to restart </li></ul></ul>
  64. 65. Operational Results
  65. 66. Lesson 1: Reduce Need For Trained Staff <ul><li>Reduce downtimes through better power </li></ul><ul><li>Nodes reboot reliably </li></ul><ul><li>Data collection and monitoring </li></ul><ul><li>Remote management </li></ul><ul><li>Tackle lack of connectivity via backchannels </li></ul>
  66. 67. Lesson 2: Simple Re-design Is Often Enough <ul><li>WiLDNet MAC </li></ul><ul><li>Improved Low Voltage Disconnect </li></ul><ul><li>External HW watchdog </li></ul><ul><li>Low-cost Power Controller </li></ul>
  67. 68. Lesson 3: Real Cost of Power Is Cleaning <ul><li>Grid power is bad quality </li></ul><ul><li>Cleaning it up is costly and recurring </li></ul><ul><ul><li>Need UPS, battery replacements </li></ul></ul><ul><ul><li>Upto 20X the monthly cost of grid power! </li></ul></ul><ul><li>Solar becomes more competitive and viable </li></ul><ul><ul><li>Equal to cost of UPS and batteries </li></ul></ul><ul><ul><li>And gives us the ability to generate power in remote areas </li></ul></ul>
  68. 69. Thank you! <ul><li>Acknowledgments: </li></ul><ul><li>Ermanno Pietrosemoli, Venezuela </li></ul><ul><li>Alan Mainwaring, Intel Research </li></ul><ul><li>Matt Podolsky, Intel Research </li></ul><ul><li>AirJaldi </li></ul><ul><li>Aravind Eye Care System </li></ul><ul><li>Resources: </li></ul><ul><li>http://tier.cs.berkeley.edu/wiki </li></ul><ul><li>[email_address] </li></ul>
  69. 70. Is it sustainable? <ul><li>Vision center setup cost: $4000 (once) </li></ul><ul><li>Cost of Link: $1000 (once), towers $2000 (sometimes) </li></ul><ul><li>Operating Cost: $150/month </li></ul><ul><li>Break even </li></ul><ul><ul><li>Assuming $10000 capital cost per center </li></ul></ul><ul><ul><li>About 2.5 years </li></ul></ul><ul><li>Next Step: 50 Vision Centers </li></ul><ul><ul><li>Implies 50,000 patients with significant vision improvement </li></ul></ul><ul><ul><ul><li>~ 500,000 other patients with lesser issues </li></ul></ul></ul>
  70. 71. The Wireless Hypothesis <ul><li>Wireless infrastructure is the first viable infrastructure for rural areas </li></ul><ul><ul><li>Much lower cost than rail, water, electricity, roads </li></ul></ul><ul><ul><li>(ignores dirt roads and mandates, e.g. aqueducts) </li></ul></ul><ul><li>… but it can lead to the rest </li></ul><ul><ul><li>By improving health care, education, government </li></ul></ul><ul><ul><li>And by creating rural income. </li></ul></ul><ul><ul><li>And particularly WiFi </li></ul></ul>
  71. 72. Schools in Batanes Prof. Manuel Ramos Jr. UP
  72. 73. SmartPhone Diagnostic Device <ul><li>Idea: collection of very simple sensors that connect to the audio jack of a “smart phone” </li></ul><ul><ul><li>Easy: heart rate, respitory rate, temperature </li></ul></ul><ul><ul><li>Medium: blood oxygen, ECG </li></ul></ul><ul><ul><li>Hard: blood pressure </li></ul></ul><ul><li>Goal: *much* cheaper, easy to use diagnostic device </li></ul><ul><ul><li>Phone converts raw data into readings </li></ul></ul><ul><ul><li>Can also forward raw data over GSM </li></ul></ul>

×