malariacontrol.net

» see malariacontrol.net as it lived, on the wayback machine

~ the obituary ~

I ran SETI@home on a family Pentium III in Ankara around 2000, which is written down elsewhere on this site. The part nobody mentions about that habit is that the client turned out to be general purpose. BOINC came out of the same Berkeley lab a few years later and would attach itself to any project that spoke the protocol, and malariacontrol.net spoke it from 2006. I never pointed a machine at it. The project looking for aliens was the one that got written up in the computer magazines.

The Swiss Tropical Institute in Basel had a malaria modeling group and a computing problem. Their simulations were stochastic, which means you build a population of fifty to a hundred thousand people and push transmission and clinical course through it. Then you do it again a few thousand times, because a single run tells you nothing. Work shaped like that splits into independent chunks almost for free, and in the middle of the 2000s buying a cluster to chew through it was expensive in a way that asking strangers for idle CPU time was not.

CERN built the wrapper around it. Africa@home was a CERN project with the Swiss Tropical Institute and the University of Geneva attached, funded through the Geneva International Academic Network, and it went public on 13 July 2006 with malariacontrol.net as its first project. It remained its only one. A closed beta had been running since at least the start of that year with account creation capped around five hundred people, and by 2014 the registration counter read 200,749.

The site itself was never designed. It was a stock BOINC install on an HTML 4.01 Transitional page with one stylesheet and a heading tag containing the words malariacontrol.net. There is no logo at the top of this grave because there was never a logo to go and find.

~ the case file ~

~ what the work units were ~

The model was OpenMalaria and it is still sitting on GitHub under SwissTPH. What came back off the volunteer machines was simulated epidemiology rather than raw data: how a bed net program plays out across fifteen years in a given transmission setting, or what an intervention costs per case averted.

The clearest use of it was RTS,S, the vaccine sold as Mosquirix. The European Medicines Agency issued a positive opinion on it in 2015, which left the WHO holding the question of whether and how it should be deployed across endemic countries. Melissa Penny's team at Swiss TPH published country-specific predictions of the vaccine's public health impact, built on eighteen months of Phase III follow-up data and on simulation models. Those simulations had been run on other people's computers. Somewhere in the chain between a volunteer's idle laptop and a WHO policy recommendation for sub-Saharan Africa there is a joint that nobody has ever bothered to photograph.

~ the last post ~

Two news items carry the entire ending. On 28 January 2016 at 09:12 UTC the team posted a note titled "Why there are no work units at the moment", apologizing for the silence. They had run into technical problems fitting improved transmission models to field data and had shut off work-unit generation while they dealt with it, expecting a short gap. Then, in their own words, over Christmas and the New Year it was difficult to assess exactly what was going on. The team members who could have addressed it were unavailable for personal reasons and there was no spare funding to hire anyone new. They wrote that it would probably be another month or two.

The second item is dated 21 June 2016 at 14:34 UTC, titled "Status and plans as of June 2016", and it says the thing straight. Volunteer computing was no longer the most efficient way to get the cycles, and the staffing to maintain a BOINC server was no longer in the building. High-performance computing had moved a long way since 2006 and institutional arrangements now gave the modeling group access to infrastructure that had not existed when they started. The investment needed to upgrade the malariacontrol.net-specific parts of the server could not be justified.

Neither post uses the word closed. The June one ends by saying the server would be kept running for now, in case some future application made this the best way to run simulations again. The last line asks participants to please keep their BOINC clients connected to malariacontrol.net. The homepage was still there in October 2016. By 28 November 2017 the domain was answering with the stock Apache page that reads "It works! This is the default web page for this server." Today malariacontrol.net has no DNS record at all.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »