Fighting malaria by accident
07 September 2006
© CERN Courier
Like several other novel projects I've worked on in the past, this one started "by accident". One day in 2005, colleagues in the communications team at CERN's IT Department were approached by Informaticiens Sans FrontiÃ¨res, a non-governmental organization (NGO), for advice on setting up a Grid or distributed computing project that could be showcased at the World Summit on the Information Society (WSIS) being held in Tunis in November that year. For example, could an "African Grid" help to bridge the digital divide? This idea fitted in well with CERN's own efforts for the WSIS, and the director general was supportive.
Having spent the last few years of my CERN career managing part of CERN's first Grid project, plus 20 years or so helping developing countries advance their Internet know-how, I realised this would be a very tall order indeed. But by coincidence, I had been working since 2004 with some students on a project called LHC@home, which involved a simple and democratic Grid-style technology called BOINC (see CERN Courier April 2005 p18). This allows people to volunteer their spare PC computer power to help design the Large Hadron Collider.
Experience with LHC@home had convinced us that volunteer computing projects could be set up with relatively modest means on the BOINC platform. LHC@home now has 30 000 computers contributing to beam stability simulations, but it is also a powerful outreach tool. It has helped create a lively online community of thousands of volunteers around the world who are keenly interested in how the LHC is progressing. One of the fascinating things is that the volunteers receive a virtual form of credit for the processing they do, and they compete between each other for who can get the most credit! Such plentiful and cheap computing resources could be an attractive reward for deserving projects and researchers in developing countries.
Convinced that we could do something equally useful and appealing for the WSIS, we went in search of resources. Thanks to a small grant from the Geneva International Academic Network (GIAN), a student team set up the new project, called Africa@home, at CERN last autumn. The team involved three computer-science students from the universities of Copenhagen, Geneva and Basel, and two young African researchers, all working under my technical supervision.
Africa@home is a website conceived and coordinated by CERN that can host a range of volunteer computing projects for Africa. The first application is MalariaControl.net, a computer-intensive simulation program developed by researchers at the Swiss Tropical Institute (STI) and adapted by the CERN-based student team. It began production in July and currently harnesses more than 6000 volunteer computers.
Malaria is responsible for about a million deaths each year in sub-Saharan Africa, and is the single biggest killer in children under five. The MalariaControl.net program simulates in detail how malaria spreads through Africa. Running the simulations on thousands of volunteer computers enables researchers to better understand and improve the impact of introducing new treatments. Speaking about the results obtained so far, Tom Smith of STI said that it has really opened up new scientific horizons for his team - they have already done more epidemiological modelling in a few months than they could have achieved on their own computer cluster in a few years.
A key objective of the project was to involve African academic institutions in the software development. It was thanks to the efforts of two local NGOs, ICVolunteers and Informaticiens sans FrontiÃ¨res, that two researchers from the University of Bamako in Mali and the Agence Universitaire de la Francophonie in Bamako and in YaoundÃ©, Cameroon, were able to join the project team based at CERN.
The server hardware was obtained by resurrecting several obsolete CERN-IT disk servers, literally from the "cemetery", where such equipment waits for destruction. Three useable servers were recuperated after considerable work, again an excellent training exercise. After initial tests at CERN the production server was installed at the University of Geneva, a development server was sent to the STI and a spare remained at CERN.
By autumn 2005 the system was ready for tests with a small group of experienced BOINC volunteers. The project was presented in public for the first time at the WSIS meeting in Tunis, where it attracted much attention, including a radio broadcast on the BBC World Service. More extensive testing began early in 2006 with 500 "typical" volunteers, and improvements and extensions to the malaria model being continually incorporated by STI. In this test phase of several months, Africa@home was able to run simulations equivalent to 150 years of processing time on a single computer. When a press announcement of Africa@home was made in July the response was phenomenal: within only a few days we passed from 500 to nearly 3000 volunteers and more than twice that number of PCs. This is currently enough for the project to handle, though we should open up for several thousand more volunteers in the autumn.
The GIAN foundation has just awarded another grant to the Africa@home partners to adapt other applications of significance to Africa to run on volunteer computers, including epidemiological modelling of diseases such as tuberculosis and HIV/AIDS. The World Health Organization will also be helping. This extended project will train technical staff at African universities to manage the servers that run the volunteer projects, and help African researchers create their own similar projects. Apart from BOINC's obvious ability to harness "free" computing power from richer countries, I believe that mastering modern open-source software such as BOINC is an excellent and practical way to train and motivate students and researchers in developing countries.
Posted: 2007-8-19 Updated: 2007-8-19