September 19, 2012
Bridging the Digital Divide
People of rural Africa understand the Internet can open doors for community building, health care, and education. To fix sluggish connections in remote areas, UCSB researchers are designing wireless network technology that uses castoff radio frequencies in a new way.
The red dirt road to Macha floods during the summer rainy season, isolating the small village from Choma, the nearest town in rural Zambia. About two-thirds of this 50km road was paved in recent years, but never completed. Before that, the journey through the African countryside required several hours by Jeep, or a chartered plane that lands on Macha’s red dirt runway. One of the first things a visitor to Macha notices is the 40-foot shipping containers planted among simple buildings. Of the three vivid blue but rusting cargo containers in Macha, only one is currently functional. Wooden doors have replaced metal ones, revealing an interior of fluorescent ceiling lights and padded cubicles reminiscent of any modern office.
Inside is a small miracle of innovation: an Internet “cyber café” where every computer is hooked up to a local wireless network. Here in one of the most rural areas of Africa, people in these shipping container Internet cafés are participating in the global network of ideas and communication through the World Wide Web. The Internet is as important to them as in any Western community, and so is their connection speed.
Between an overworked uplink connection and supersized websites, Internet connectivity in Macha is achingly slow. The problem is two-fold, explained Professor Elizabeth Belding, a Computer Science professor at UC Santa Barbara. The Internet should be opening doors for rural users to access the same popular social networking sites everywhere else in the world, but those sites are jamming the network. “The entire community is connected through a slow satellite uplink that operates at about 1Mbps, which is slower than the DSL or cable modem connection most people in the U.S. have in their homes,” said Belding. Her studies of Internet usage in rural Africa have shown that, more than any destination on the Internet, Macha’s people are Facebook power users. According to a study by Belding and her team, between 15 and 20% of all HTTP requests from Macha are to Facebook.
Traffic for Facebook has to bounce up to a satellite uplink, down to a server across the world, back up to the satellite, and finally back to Macha — all to send one wall post to a friend in a nearby city or down the road. Users’ patience with this redundant system is beginning to fray.
“People have experienced the Internet enough to understand the benefit it can bring,” Belding explained, “and they are extremely frustrated by the long waits and high rate of aborted sessions. That’s unfortunate because they have the most to gain from the wealth of information available on the Internet.”
That limited access is why Belding’s research partner, Professor Lisa Parks of Film and Media Studies, and several graduate students are traveling to Macha this summer to help build a new Internet connectivity system — an integrated program known as VillageNet.
To understand Macha’s Internet problem, first imagine it’s 1997. Even a freshman in Professor Belding’s Introduction to Computer Communication Networks class remembers what it was like trying to access the Internet fifteen years ago: the shriek and clang of the dial-up modem, the glacial downloading speeds, and the stripped-down web sites. But while the process of connecting then was slow, Web surfing was smooth, because websites were fairly uncomplicated at an average size of about 14KB. By 2011, websites were running closer to 679KB — which is 48 times the size of a website in 1997.
Most of the world hasn’t noticed this ballooning, though, because as websites got bigger and more sophisticated, Internet connectivity got faster. The clunky dial-up modem was upgraded to DSL, then cable modems, and download speeds kept up with the exponential growth of website size. That accelerated speed is the reason most people in the first world can view a Facebook page, join a hangout on Google+, or share files on Dropbox.
The village of Macha connects to the Internet by way of a 2Mpbs link to an Internet Service Provider. Actual connection speeds rival those of an old dial-up connection from fifteen years ago. Internet service is not only much slower than the cable connections used in the global West, it is also exponentially more expensive. The community pays between $3,600 US per month for the equivalent of a dial-up system — a steep expense for a village where people make an average of one dollar per day.
“...At any time of day, up to three hundred people are trying to access the Internet on a connection no more powerful than a single modem was two decades ago.”
People make camp outside the hospital in Macha waiting for their family members to be discharged.
View image gallery below
The hospital, the school, and the cargo container Internet cafés all share a wireless mesh network — a system that is frequently crippled by a single user trying to upload media or download one large file, such as a PDF or a YouTube video. According to
Belding’s research and personal interviews conducted by her doctoral student Veljko Pejovic in 2010, the top 0.1% of the largest traffic flows on the local mesh network holds up more than 50% of the bandwidth shared by several hundred people. Because of this cramped bandwidth up to 75% of uploads fail in Macha, which makes simple tasks both infuriating and impossible.
Similarly, large downloads are frequently abandoned — either by the server or the user — because they are taking too long to load. But by the time the download is aborted it has already jammed the network. One YouTube video about a honey badger, or a PDF of a medical journal article, is all it takes to cause the network to crash and burn.
Belding and Parks, funded by a $1.2 million grant from the National Science Foundation, are building a system to address exactly this problem. Together with Gertjan van Stam at Macha Works, a program that creates community-initiated solutions that re-invest in the community, Belding and Parks hope to build a socially-informed local network that serves the village and the surrounding area.
VillageNet has three components: VillageLink, VillageCell, and VillageShare. Each component is designed to achieve the team’s goals of keeping local Internet traffic local, and to shift large uploads or downloads to periods of decreased traffic. The three elements of VillageNet are intended to bring Internet connectivity to the community of Macha — and, in the future, other remote communities just like it — through both new wireless technologies and network architectures.
Each of the three components of VillageNet solves a specific problem. VillageLink is intended to help alleviate some of the congestion of the wireless connection. It utilizes the white spaces spectrum, frequencies between 50 and 700 MHz, that were originally used for television broadcast before it went almost completely digital in recent years. These extra bands have the added advantage of good “foliage penetration properties” — ideal if you are trying to build a wireless network in a village full of trees.
“It is challenging to cover areas with very sparse population. That is why we need to develop a technology that can provide wireless links over long distances,” commented Mariya Zheleva, one of Belding’s PhD students who spent the summer of 2012 in Macha.
VillageCell will provide a low-cost cellular network, providing free local voice and SMS communication. Cell phones in Macha are only usable in the town square, and there are no landline phones in the surrounding region. Reliability is also an issue, added Zheleva, “When I was in Macha, the commercial cellular network was down every other day for a few hours.”
To address both these problems, VillageCell will include a localized cell phone system that will use OpenBTS (open base transceiver station) to intercept cellular signals and convert them into VoIP (Voice over Internet Protocol) packets.
In addition, VillageCell presents the opportunity to implement ImmuNet, a public health impact project by Belding and fellow Computer Science professor Amr El Abbadi. ImmuNet works on the VillageCell architecture, and keeps track of the vaccination status of a particular population.
Professor Elizabeth Belding (far right) and her computer science research team, left to right: Arghyadip Paul, David Johnson, Mariya Zheleva, Maggie Weng, and Veljko Pejovic
Professor Amr El Abbadi (left) and graduate student Ceren Budak
Professor Lisa Parks (left) and graduate student Abigail Hinsman.
El Abbadi and Belding received a Grand Challenges Explorations Award for $100,000 from the Bill and Melinda Gates Foundation for ImmuNet, a program that uses cell technologies to improve vaccinations in rural populations. ImmuNet is an information distribution system that contains an underlying database called VaccStore. The database helps doctors and families stay connected with one another, both by keeping track of an individual’s vaccination record and by providing a link between doctor and patient.
“The lack of technology makes the work of health care providers on the ground very challenging,” commented El Abbadi. “Our goal is to build an integrated communication and data storage system that maintains current health records and helps notify patients when they are due for vaccination, or when they travel to a highly infected region. The potential is very promising and exciting.”
In Macha, a technology such as ImmuNet could help solve problems like keeping a child’s vaccination history available and sending reminders for important booster shots. Local health care providers have almost eradicated local spread of polio, a viral disease that leads to paralysis, and now they want to extend their reach to eliminate other potentially deadly diseases. ImmuNet is an incredibly powerful mechanism that could help achieve this goal, and all it requires is a cell phone.
VillageShare is a time delayed proxy server designed to improve local traffic on the Internet. It has two components. The first is a time delay proxy that intercepts large file uploads that might put a squeeze on the bandwidth — images or videos, for example — and delays the upload until Internet traffic thins down for the day. Most of Macha’s computers are in public places, such as the Internet cafes or the schools, so there is more bandwidth available late at night. The second is a file sharing component that stores files locally, so that two local users can share content without needing to upload it to a remote server. Using a local server reserves the satellite bandwidth for downloads that require the Internet.
Ideally, Belding explains, these three components will help make cellular communication and Internet connectivity ubiquitous throughout the area of Macha. “We want residents to have Internet access in their homes, offices, schools, and public buildings,” Belding says.
Faster Internet in schools could mean a world of difference for local children. “Imagine going to a school with only a handful of books in the library,” Belding suggested, “and then suddenly having the information on the Internet available to you — what a difference that could make.”
Schools are an especially important consideration, because nearly half of the local population is under 12 years old.
Faster Internet in the local hospital means doctors can easily reorder medicines or consult with experts remotely. Farmers can research crop rotation strategies or new markets for their harvest. People in neighboring communities can exchange information and news helpful for daily living.
A faster connection also means that family and friends can keep in touch with people both within and outside the village through Facebook and Skype, which are popular in the community.
For Belding, Parks, and their team the real innovation of VillageNet isn’t the wireless technologies, it is the people of Macha. Development projects in villages like Macha often fail because they are conceived of and structured with no input from the community the project is intended to help. “If you don’t have buy-in from the local community,” Belding said, “and if you don’t understand the needs and wants of the community, the solutions you design are destined for non-use.”
The ethnographic component of VillageNet is what makes the project unique, and hopefully what will make it successful. While there are hundreds of information and communication technology (ICT) development projects going on in countries around the world, VillageNet is one of the few that includes an ethnographic element. VillageNet’s intention, Parks explained, “is to foster dialogues between community members and those designing the network.”
The collective goal of the VillageNet team is fused with the aim of Macha Works, which is that “the only people that can develop Africa are people from Africa itself.”
With that in mind, said Parks, “It is important that the network be designed in a way that is sustainable, and not simply installed without any community input or involvement.” The best way to understand what a community needs is to ask members of that community “what do you need?” And that is why Parks conducted 60-minute, qualitative interviews with at least thirty Macha locals this summer. She will keep in touch with those same people in order to access how well this new network architecture has worked for the community.
Parks and her graduate student, Abigail Hinsman, started their ethnographic study of ICT use in Macha this summer. They conducted what Parks describes as a “radial ethnography”: they recruited ten community members to be their video partners, and those video partners conducted interviews, asked useful questions, and uploaded videos of their interviews with other locals onto the server. These partnerships are actually a crucial component of the ethnography, Parks said, because it “is designed as a collaborative approach. We’re trying to break down the divide that’s characterized conventional ethnographic methods, between the ‘Western expert’ and the ‘racialized other.’ We want the people we’re studying to participate, evaluate, and benefit from the research being conducted.”
“Dr. Parks’ component is critical,” said Belding. “Her study can help make sure these technological solutions that are developed will be adopted by the community.”
This is only the beginning of Belding and Parks’ research. Over the next three summers, Parks will reconnect with these same video partners and the locals they interviewed, and conduct new videotaped interviews that chronicle their Internet use and habits, their feedback on VillageNet, and how they feel Internet connectivity has impacted their community. Belding’s team will also return, implementing components of their technologies, testing them, and checking in with residents on how these changes have affected their Internet use. Their hope is that, by the end of the four-year grant, the difference in Internet connectivity will be like fast forwarding from 1997 to 2012.
As excited as Belding, Parks, and their graduate students are to make a difference in Macha, they hope that Zambia is a beginning rather than an end. As they perfect the network and its use, their expectation is that there will be many more iterations of VillageNet around the world. “Ultimately,” Belding explained hopefully, “our goal is that our research is useful widely, and it is applicable to any remote community, anywhere in the world.”