Le 20 et 21 Mars 2011 aura lieu une serie de conferences, dont pendant 2 jours, sur BOINC !
Il s'agit de Asia@home, qui se deroulera a Taipei (Taiwan). Evidemment je compte sur votre venu !! (Un peu loin certains diront )
Enfin pour ma part j'irai aux conferences. (Le 21 est sur invitation, je vais tacher d'en avoir une, si je peux me liberer du travail)
Voici le manifest: (en anglais, mais on retrouve David Anderson....d'ailleurs je sens que je ne vais pas pouvoir stopper mon envie de lui demander pourquoi il supporte le projet "The Charity Engine" !!! , il y aura aussi le LHC, QCN, Climateprediction, etc...)
This Asia@home hackfest focuses on applications of citizen-based volunteer computing and volunteer sensing to earthquake science in South-East Asia. The purpose of the hackfest is to test new concepts in distributed software and hardware, and brainstorm about how to increase web-based public participation in earthquake science.
The two-day hackfest tackles two specific challenges:
The first day focuses on integrating large-scale simulation packages for earthquake science with the volunteer computing platform BOINC (http:// boinc.berkeley.edu/), so that home PCs and laptops may contribute to simulation-based research.
The second day focuses on how to deploy the volunteer-based earthquake sensing project Quake Catcher Network in Taiwan and neighbouring regions of South-East Asia, in a way that also contributes to public education about earthquake science.
Day 1 (Sunday 20 March): Citizen cyberscience and earthquake simulation.
Context: highly parallelized code for earthquake simulations is normally optimized to run on tightly coupled processors, and requires very large (>1GB) input and output files. Volunteer computing represents a huge and essentially free source of processing power, but is usually only appropriate for embarrassingly parallel applications, with limited I/O. However, constant progress in consumer hardware, software and bandwidth to the home offers new opportunities to adapt earthquake simulation code to volunteer computing. This could provide researchers in SE Asia with a powerful, low-cost computational platform for earthquake research.
9:00 "Shakemovie@home" Li Zhao, Institute of Earth Sciences, Academia Sinica
We are now computing a database of 3D strain Green tensor (SGT) for Taiwan (100TB+) so that once a CMT is determined a program will be triggered to retrieve all the Green's functions needed to generate the synthetics to make shake movie and shake maps. The SGT database will have around ~7000 grid point (~64x110 covering all Taiwan area by 25km distance for adjacent two points). Each grid point will have about 15GB SGT data and is independent of any others. When there is an earthquake, we need to extract from each point the SGT for the earthquake location and calculate the seismogram for that point. In this talk, I will describe how the SGT database generation will be done by the Service Grid while the extract part (to calculate the seismogram and create the shake map/movie when there is earthquake happens) could be done by the volunteer computing model
9:30 "The limits of volunteer computing" David Anderson, Space Science Laboratory, UC Berkeley
In this talk, I'll cover: statistics on the hardware characteristics of volunteer hosts (CPU capacity, RAM, free disk, GPUs, network BW); BOINC's mechanisms for matching jobs and hosts; BOINC's mechanisms for dealing with extreme requirements; (RAM, disk, runtime); how we handle MPI apps in BOINC.
10:00 Coffee break
10:30 "Lessons learned from running large-scale simulations on a volunteer platform" Carl Christensen, School of Earth Sciences, Stanford University
In this talk I will describe how we successfully distributed large-scale climate simulations to volunteer PCs, in the project ClimatePrediciton.net from Oxford University. I will also describe first efforts to integrate MPI with BOINC in a way that could be useful for the earthquake science community.
11:00 "LHC@home and CernVM – a new approach to porting large-scale applications to BOINC" Daniel Lombrana Gonzalez, Citizen Cyberscience Centre,CERN, Geneva.
In this talk, I will review a new application in the volunteer computing project LHC@home, called Test4Theory, which is currently being alpha tested at CERN. This application uses virtual machine technology and a range of other software tools to handle large, complex software environments that are rapidly evolving, so that they can be distributed using BOINC.
11:30 Brainstorming session – planning teams and activites for the afternoon.
13:00 – 18:00 Hackfest
The goals of the afternoon session include:
Running existing seismological simulation code on BOINC;
Testing performance on various multicore machines;
Exploring porting to GPUs and other processors;
Setting up a campus grid for seismological simulation with Jarifa.
19:00 Hackfest Dinner at the Famous Tea Restaurant
Day 2 (Monday 21 March): Volunteer sensing for earthquake science in SE Asia
Context: The Quake-Catcher Network (QCN) is a collaborative initiative for developing the world's largest, low-cost strong-motion seismic network by utilizing sensors in and attached to internet-connected computers. QCN uses BOINC as a platform for communication between distributed sensors on the Internet and a central server. The project has attracted considerable public interest, and is used by some schools to educate children about earthquake science. Although the sensors have limited sensitivity, their ubiquity can provide researchers with new insights into earthquake dynamics, and their low-cost makes them particularly interesting for deployment in developing regions of SE Asia, where sophisticated sensor networks may be lacking.
9:00 "Introduction to the status of earthquake detection in Taiwan" [TBC]
9:30 "Earthquake detection in South-East Asia" [TBC]
10:30 "Quake-Catcher Network: the sensors, the science and the outreach" Elizabeth Cochran, Department of Earth Sciences, University of California, Riverside.
In this talk I will introduce the Quake-Catcher Network and outline the current status and recent results of the network. I will provide an overview of the client- and server-side algorithms for earthquake detection, including recent tests of the network following the recent earthquakes in Chile and New Zealand. The past, current, and future sensor models will be described, including shake table test results. Finally, I will describe our experiences recruiting and maintaining volunteers for the project with and emphasis on classroom outreach.
11:00 "Quake-Catcher Network: the client and the server" Carl Christensen, School of Earth Sciences, Stanford University
In this talk I will review how the Quake-Catcher Network (QCN) works on the volunteer client machines (laptop & USB drivers for the sensors) under BOINC. QCN is quite different from other BOINC projects, as it is very low-CPU but requires fast turnaround of reporting for seismic events over the Internet (ideally within a second). I will also discuss the various server-side components and modifications to BOINC to handle our unique requirements. This includes the "instantaneous" triggers from seismic events via the BOINC trickle mechanism, matching host IP addresses with geo-location (latitude/longitude lookup via GeoIP and user input), potential earthquake detection and possible early warning system, data uploads, real-time processing and post-processing on the server. Other related QCN projects, such as our continual monitoring system and building shake studies will be presented, as well as ideas for data redundancy of the database and seismic triggers of international scope.
11:30 Brainstorming session – planning teams and activities for the afternoon.
The goals of the afternoon session include:
Installing and testing QCN USB sensors;
Setting up a QCN server;
Developing a web-based dissemination campaign for QCN in Taiwan;
Developing a strategy for regional deployment of QCN in SE Asia.
18:00-19:00 Summary, Future Plans and Evaluation of the Hackfest