One woman's path through doula training, childrearing, and a computer science Ph. D. program

Friday, November 11, 2011

Connecting the Disconnected: Improving Internet Access for the Other Four Billion

This post is about Connecting the Disconnected: Improving Internet Access for the Other Four Billion with Professor Margaret Martonosi from Department of Computer Science at Princeton University, at Grace Hopper Celebration of Women in Computing.

Information technology is an enabler.  It is an enabler in education - for people to learn, through online courses, articles, and things in the public domain; in health care - getting an ultrasound when needed; in agriculture - the ability to find out why your crops aren't thriving; and in open and fair government - the ability to discuss politics openly with other people.

This presentation was about C-LINK, a type of  delay-tolerant vehicular network.  (Note: This vehicular network is an example of Sneakernet, though I don't think Margaret ever called it this.)

The current state of universal connectivity

Sure, the US is connected.  Over 75% of the country has wired or wireless access.  But there's the digital divide. The digital divide means that the places that need this technology the most are the places with the most impaired access to it.

The three factors influencing universal information technology access includes relevant and accessible software, using effective and affordable hardware, and have universal connectivity.  C-LINK, Margaret's project, was influenced by all three.

The problem with affordable hardware is the "last mile" problem: it is easy to connect the majority of the people, but in the last mile, you have to extend connectivity to rural or hard-to-reach regions.  This can be complicated, and costly.  In rural Africa (for example), the last mile is of a whole different scale.  So then we consider wireless technology, which is expensive, we need to think about how it still needs a wired backbone, requires ongoing maintenance, is subject to corruption or salvaging (because copper sells), and it needs political support.  On the other hand, wireless technology is leapfrogging wired connectivity (especially in developing regions).  Mobile and cellular is a big growth area and are increasingly penetrating the world.  But even though there is so much cellular technology going on, that doesn't mean that it's cheap.

Let's look at effective and affordable hardware.  Although there are efforts to build cheap hardware (such as the hundred-dollar laptop made by the One Laptop Per Child organization), it's not taking off.  It's just not pervasive -- most of Africa and Asia have less than 10 computers per 100 people.  That's even when you consider microcontrollers as "computers."  So what is the world's computer?  The world's computer is a cell phone.  About half of the world's adults own one.  There are more cell phones in India than credit cards -- and cell phones, even when they're not smartphones, are chock full of interactivity.  (Note: Can you imagine an accessible Qwiki for developing nations that works over SMS?  I can.)


Wired connectivity: availability is increasing and costs are dropping.  For example, transit prices in Kenya have dropped to $120/Mbps -- which are similar to US prices in 2003.

Delay-tolerant network: Imagine a big city with a broadband connection, and a nearby village.  Imagine that the village is on some rural bus or taxi route to the city.  Now, imagine that the bus has a laptop inside, with a large hard drive and a wireless access point.  People (and goats and chickens) board the bus in the village, send some requests on the laptop.  Although the laptop isn't connected to the broadband, it can queue the request and send it when it reaches the city, and bring back the reply when it returns to the village.  If the bus visits the village every 5 hours, that's a 5-hour latency -- but at least it is access to information.
Never underestimate the bandwidth of a station wagon carrying tapes hurtling down the highway. -- S. Tanenbaum 

The vehicular delay-tolerant network is very low-cost, easy to deploy, and has a very high data carrying capacity.  Villagers can collaborate on their queries, and the results can be available to others in the village doing similar web searches (called collaborative caching).  The system can also be improved by clever prefetching -- the computer on the bus can be thinking hard over the next several hours about what else can be proactively fetched based on the current queries.  Clearly, prefetching all of the links on a page would be smart, but maybe so would translating a page from a local language to a more commonly-used language.  This isn't a microcontroller with a tiny cache.  This is a big hard drive.  If the hard drive is huge, there is no worry about performance loss in prefetching things that will not be used, but the overhead of not having things that may be useful is huge.

"Come back tomorrow. Your data will be on the bus."

The authors designed, built, and tested C-LINK, the vehicular delay-tolerant network, over a week in Nicaragua.  The city was Somotillo, Nicaragua -- no skyscrapers, but there was a school with a computer cluster and wired connectivity and a small, expensive Internet cafe with 2 computers.  Participants were school children, invited to come in and browse the highly-delayed web.  About 80% of the participants in the project had never used the Internet before.  One of the cool things that the project looked at is when participants sent at the same query, or searched collaboratively.  With each successive trip into the village, the bus brought back more data.  As the village's cache filled up, users' miss rates plummeted.  Data were on their computers or on the computers of their collaborators -- remember collaborative caching?  The authors found that highly correlated access patterns provided strong incentive for collaboration.  Oh, and we can evict old data or data that has not been accessed recently.  Any cache eviction policy will work here.

Then Margaret and her team considered using laptops and other mobile computing devices and, despite the worry of data loss, found that going mobile wasn't so bad.  If anything, it made for more kiosks -- little hubs for collaborative information.  Then, they found that the requests for data were fairly small, so it was possible to exploit cellular connectivity, where available.  They looked at data brought back by the once-a-day trip into town, by the 5-times-per-day bus ride, and a hybrid data retrieval mode by combining these with exploiting SMS.

Look around you

Other software that the authors found interesting and noteworthy included these.

  • TEK is an e-mail-based web browser.
  • M-Profesa helps Kenyan children prepare for the secondary school test through SMS.
  • Ushahidi, also originally Kenyan, helps with crowd-sourcing of information.

Getting more involved

Imagine what you could do if you could alleviate teacher shortages by having better distance learning technologies.  Imagine what you could do if you could have better information flow and expose corruption in the government.  Get involved in Engineers Without Borders and attend Development conferences, such as ACM DEV 2012.  Make a company.  Build stuff!

No comments:

Post a Comment

Related Posts Plugin for WordPress, Blogger...