Evolution@home: observations on participant choice, work unit variation and low-effort global computing

摘要

Many simulation efforts in ecology and evolutionary biology employ individual-based models that are well suited for including many biological details. These models often pose serious computational challenges if all biologically interesting parameter combinations are to be explored. The challenges are even greater for biologists who often lack supercomputing facilities and the manpower for implementing complex global computing systems such as SETI@home. Under such limiting conditions, evolution@home started as a one-man effort to distribute simulations of Muller's ratchet to Internet-connected computers of participants from the general public. This paper addresses experiences in low-effort global computing made with evolution@home over more than four years. It shows how allowing participants to choose the class of computational complexity they want to contribute to can help to deal with the bewildering variety of computational complexities that easily result from individual-based models. Results suggest that, as a first rough approximation, participants' complexity choices are distributed surprisingly even over all reasonable classes of CPU-time and RAM requirements. More often than not, participants tend to finish the simulations they start, if they are committed enough to submit any results at all. Potential uses of intermediate simulation results are discussed and the error of magnitude is introduced to help to deal with imprecise CPU-time predictions. Experiences with the choices of over 300 users who have contributed more than 100 000 simulations with a total of over 80 years CPU time are reviewed.