Conveners
Supporting EGI users with crowd computing based on Desktop Grids
- Robert Lovas (MTA SZTAKI)
- Ad Emmen (EDGI, DEGISCO, e-IRGSP3 projects)
Description
The International Desktop Grid Federation is working closely together within an EGI Virtual team to make it as easy as possible for scientists to use Desktop Grid resources: unused computing capacity from university computers (local desktop grids) or from volunteer citizens at home (Crowd computing).
The session presents the results to scientists and community organisers:
16h30 Opening & Welcome - Arie Vleugel, Member of the Board, Stichting IDGF
- First we concentrate on how to use computing resources from Desktop Grids while staying in the comfort zone of EGI.
Use the EGI connected portals, job submission and monitoring tools to get access to an additional thousand or ten-thousand processors. We will also explain to community organisers what is needed to get their community "Desktop Grid ready". EGI and IDGF have formed a Virtual Team that can assist.
16h30 - 16h45 The HADDOCK WeNMR portal goes crowd computing - Alexandre M.J.J. Bonvin, WeNMR, Faculty of Science - Chemistry, Utrecht University, the Netherlands
The HADDOCK web portal is a widely used scientific portal for the modelling of biomolecular interactions. It makes use of the EGI grid infrastructure for job submission. Users interact through a user-friendly web interface. Each user submission translates into several hundreds of individual grid jobs that are handled by the complex workflow beyond the portal. To date HADDOCK counts over 3900 registered users worldwide and has resulted in the last year in over 860.000 grid jobs for a total of over 150 CPU years (normalized CPU time (kSI2K) - source EGI accounting portal).
Until recently, the grid-enabled portal was only making use of standard gLite-based submission and retrieval of jobs via a user-interface. In collaboration with the IDGF team we have ported and validated the HADDOCK computational engine, the CNS software, into a Linux virtual machine for BOINC. This effectively enables the HADDOCK portal to send jobs to IDGF resources using the same gLite-based mechanism as for regular EGI resources.
The IDGF-enabled HADDOCK portal has now been in operation since a few months, sending about 10% of its jobs to IDGF resources. Results show that the Desktop Grip computing resources are performing at a similar level as regular EGI sites.
URLs:
http://haddock.science.uu.nl/enmr/services/HADDOCK
http://www.wenmr.eu
16h45 - 17h00 Spatially explicit ecosystem modelling with the help of workflows and desktop grid technology - Peter Ittzes, BioVeL
Understanding and predicting ecosystem functioning is a key subject of ecological research. Effects of different environmental changes on ecosystems like climate change is a frequently occurring topic, even in the public media. However, modelling these systems is complex and computationally demanding due to the high number of parameters, various options and exhaustive meteorology input requirements. This problem is multiplied if modelling happens over a heterogeneous and patchy geographic landscape. In the BioVeL consortium with the cooperation of MTA SZTAKI a toolset is being developed, that integrates scientific workflow and desktop grid technology. We are working on integration with GIS for spatially explicit ecosystem modelling, which utilizes Taverna workflows, EDGeS desktop grid and Open GIS technology. Main steps of Desktop Grid integration will be demonstrated in the presentation.
17h00 - 17h15 Utilisation of IDGF computational resources: alternatives - Jozsef Kovacs, MTA SZTAKI, Hungary
The IDGF-SP infrastructure consists of volunteer and institutional BOINC projects, bridges to and from EGI (European Grid Infrastructure) type of grids (e.g. gLite), cloud resources, science gateways, monitoring and accounting systems, and virtualization support for the volunteer resources. Several EGI communities are already utilizing these volunteer resources. The infrastructure provides several alternatives for accessing and submitting jobs to Desktop Grids. This presentation gives an overview about the different ways and strategies communities are using the IDGF computational resources.
- Second the session will explain how to organise a local Desktop Grid within a university or institute: connect the organization's computers into one huge computing engine.
Port applications to this engine and provide additional number crunching to scientists. IDGF has a support team that can help.
17h15 - 17h30 Establishing and operating local desktop grids - use the power of your institution's idle computers - Tamas Kiss, University of Westminster, United Kingdom
Research institutes, universities and companies all operate large numbers of personal computers to support their employees or students. These PCs are purchased by the institution for supporting its major business, education or research activity, and maintained centrally by an IT department. However, these computers are very often idle, just sitting on the desk and doing nothing. This happens during the night or out of office hours, during holiday periods, at lunchtime, or between scheduled lessons. Using desktop grid technology, for example BOINC, the free computing power of these resources can be utilised for useful scientific or business computations that require large computing capacity. This can be done without compromising the availability and performance of the computers, or negatively influencing the user experience.
The advantage of this solution is that no specific new hardware needs to be purchased, and the maintenance is also performed centrally by the institution independently from the desktop grid utilisation. This presentation will demonstrate via the example of the University of Westminster BOINC desktop grid how local desktop grid infrastructures can be set up and operated. The Westminster Local Desktop Grid connects almost 2,000 laboratory PCs into a powerful computing resource supporting diverse application areas. Porting, deploying and operating applications on these local resources will also be covered and illustrated with examples from both research and teaching. Finally, the support available from IDGF covering the full lifecycle of local desktop grid set-up and operation will be outlined.
- Third we will talk about Crowd computing.
It is exciting to see thousands of enthusiastic citizen volunteers donating computing time to your applications. Not only can you get your work done more quickly, but also it provides an excellent opportunity to spread the knowledge about your research to society. IDGF has a special, supported, crowd computing Grid dedicated to EGI applications.
17h30 - 17h45 Overview Crowd Computing - Ad Emmen, Stichting IDGF & AlmereGrid, The Netherlands
In the last years there have been a lot of initiatives labeled "crowd". We have crowd funding, crowd sourcing and also crowd computing. It all has to do with using the strength of the "crowd", the large number in a crowd, so that, although each individual contribution may be small, the whole crowd contribution can be very large indeed.
Crowd computing is about sharing computer capacity. One computer may be small, but the power of a million put together can match a supercomputer's number crunching capabilities. Although most crowd computers are smaller than that, even 10.000 computers in a crowd can already do significant computing work.
Crowd computing can not only be used for number crunching, but also for storage, and data.
There are many Crowd computers in the world, some old, some new, some small, some very large. Some of the most mature ones are the so called voluntary desktop grids that are mainly aimed at helping science. About 50 of them are member of the International Desktop Grid Federation (IDGF).
In this presentation we will give a short overview of crowd computing and a number of examples of crowd computers that can be used by members of the European science community.
17h45 - 18h00 Panel discussion and Wrap-up
More details on the programme and speakers at: http://desktopgridfederation.org/programme20
