|
Results
calculated by the simulators
As the raw data of
running simulators is reviewed, you will be able to read here, what
the results mean that you contributed to. Part of this will be in
non-technical language to allow many readers to understand. Other
parts are for scientists working in that area.
It will still take some time until a nice report can be published
here. This does not mean the results you submit are in vain. Please
bear with the inconveniences of the current semi-automated system
until the next generation framework has arrived. Your results
are important whether they have been computed by an nice looking
automated simulator with immediate web-feedback or by a tedious
manual commandline.
First
scientific
paper about
evolution@home
with IT-results is ready for download
(24 October 2002)
While biological results
are still being written up, the first scientific
paper based on results submitted to evolution@home is ready
for download now. It addresses some fundamental issues about global
computing projects that span a large number of orders of magnitude
in computational complexity.
For participants of evolution@home
it has become natural to choose the complexity of their work units.
However, this users-free-choice-system is new to global computing
and it was not entirely clear from the beginning, how well it would
work in practice. The first scientific paper about evolution@home
adresses these and similar issues, like e.g. dependence of performance
on CPU-cache structure.
Another problem encountered was to quantify, how many orders of
magnitude predicted and actual computing complexities differ on
average (for releases 1-5 of Simulator005). As the assymetric properties
of the commonly used relative error did not allow proper quantification,
the "error of magnitude" was defined in this paper. Its
values are symmetric around zero (= no error) and easy to understand
intuitively (e.g. +1 is one order of magnitude, i.e. 10 times longer
than predicted).
This paper
was presented at the 2nd
International Workshop on Global and Peer-to-Peer Computing on Large
Scale Distributed Systems at the IEEE International Symposium
on Cluster Computing and the Grid (CCGrid'2002) in the Berlin-Brandenburg
Academy of Sciences and Humanities,
May 21 - 24, 2002 Berlin, Germany (see http://www.ccgrid.org/
and http://ccgrid2002.zib.de/)
and has been published in the corresponding proceedings: Loewe
(2002) "evolution@home: Experiences with work units that span
more than 7 orders of magnitude in computational complexity",
pp. 425-431 in: Bal et al. (eds) Proceedings of the 2nd IEEE/ACM
International Symposium on Cluster Computing and the Grid (CCGrid2002),
Berlin, Germany, IEEE Computer Society. You can get it free here.
As no high-scores existed
at the time of the writing of the paper, the following acknowledgements
are added here: RabeRudi, Paranoia Retnek, arswitchman switchman
and many others made significant computing contributions as participants
of evolution@home. Without their help it would not have been possible
to conduct this work.
Successful
prototypic analysis of first evolution@home results
encourages further computations (14 September 2001)
The first glance
at the results of evolution@home is promising. Despite its tedious,
semi-automated way of operation, more than 120 individuals contributed
nearly 5 years of CPU-time from April to mid August 2001. Some of
the run-files have been removed from the website, as they are completed
now.
Until an automated priority setting system for run-file production
is in place, please help compute simulations with longer computation
times (less work for you and your modem, more work for your computer).
At 13th of August the cut for the first (still manual) analysis
session was made. All results submitted by email until then were
included in the analysis made for a poster at the Eighth Congress
of the European Society for Evolutionary Biology (Laurence Loewe,
"Predicting extinctions due to Muller's ratchet in humans and
bacteria", 20-25 August 2001, Aarhus, Denmark). Details on
the biological meaning of the results will be published in the coming
months. While it will still take some time until analysis of results
submitted will become fully automated (including high-scores computation),
a first assessment allowed to estimate some global statistics. More
than 500 results mails with more than 100 MB of results were recieved
by more than 120 participants.
The combined, sustained number crunching capacity of Simulator005
in that time was about 2 million individuals per second. This equals
to nearly 10 fully dedicated up-to-date PCs running continously.
While this seems little compared to other projects that are fully
automated and have been well advertized, this is remarkable given
the amount of manual interaction involved, the virtual absense of
advertisement and the lack of high-scores that ensure every
participant the project is still active. To answer a question many
may have thought and a few have asked: evolution@home is active
and will remain active even if it takes some time until the website
is updated. There are no plans to stop it and if this changes,
it will be announced at this website. All results submitted will
be evaluated manually from time to time and automatically once the
corresponding software has been developed. After that updates of
the website will become much more frequent.
|
