open-source Notes

Notes of an open-source programmer.
31 Aug

6th BOINC workshop London – day 1

The first day started with a talk by David Anderson summarizing the development of BOINC over the last year. Especially the new Creditsystem. Another visible addition to the manager is the new notices tab. This will show translatable information for configuring BOINC, project news and even private messages in the new generation of the BOINC manager, currently in testing. Upcoming tasks for the developers are, among other things, the rewrite of the Manager (at least the Simple View) and the integration with Drupal.
After a short coffee break Benjamin Knispel presented the search for pulsars using the Einstein@home userbase. This started in March 2009 as a side project along the gravitational wave search and found 2 new pulsars within the last weeks using the computational power of volunteer computers.
Following this, Tolu Aina showed some of the projects the group around climateprediction.net currently works on. One of them tries to collect information about the power consumption using external wattmeter devices. Another side-project is BOINClite, a lightweight BOINC client that should run one project on Playstation 3 processors.
Short before lunch Adam Bazinet introduced us to a long running but almost unknown project called the Lattice project. He points out how they combined different computing techniques such as High Performance Grid, dedicated Clusters and Volunteer Computing using BOINC to solve there biological problems.
Martin Korth concluded the first series of talks for today with a fast presentation of MonteCarlo@home. He has shown what way quantum chemistry will go in the future and what they have learned from MC@h, lots of biomolecular and quantumscience that feels like rocket science for ordinary people. There is also a new project in the pipeline that will feature a new method to compute quantum chemistry problems.
Lunch break was at the University’s cafeteria with a delicious chicken tandori.
After this relaxing break Kevin Reed from IBM presented the projects and procedures of World Community Grid. 17 papers were published in the last years based on the computing power what volunteer computers provided.
Then Nicolas Maire presented the current status of the malariacontrol.net project. The matter of malaria got a massive amount of attention in the last years and the methods for creating models for malaria control and distribution. This also led to more scientists that are working on this problems. This scientific boost also leads to a higher need for computational power where BOINC and its volunteers kick in.
A short journey into the world of mathematics was presented by James wanless who introduces us to primes and factorization and how BOINC contributes to this area of science.
The next talk was going in a similar direction. Yu-Ting Chen from Puzzle@home explained how they want to find a sudoku puzzle with a unique solution and just 16 clues. Unfortunately this is no public project as they run this on a dedicated computing grid.
The last block of talks for this day was started by Diogo Ferreira and his presentation of a Plugin-Manager for the BOINC client which allows the usage of predefined Execution Environments or virtual machines. The main problems here are the high bandwidth and disk-space requirements.
A similar idea was shown by Ben Segal from CERN. Over the last 2 years he and several summer students and experts from BOINC developed a very special virtual machine wrapper used at CERN. They developed a customized virtual machine that gets distributed as a BOINC task to the clients and runs infinite on the machines. This is ok for a controlled computing grid where network connections are close and fast but not for public heterogeneous grid like BOINC.
The next talk by Chris Reynolds was geared towards the same idea around virtual machines and predefined custom execution environments that are send along as BOINC tasks. There are not all BOINC functions build in and some are even impossible to implement but this seems to be the best promising candidate.
The last talk by Derrik Kondo showed a new mechanism to estimate the cost of running BOINC tasks on dedicated grids in order to allow a fast turnaround and completion of tasks. This has to be implemented of course.

Leave a Reply

© 2017 open-source Notes | Entries (RSS) and Comments (RSS)

GPS Reviews and news from GPS Gazettewordpress logo