Post-doc position in Grid computing
|
|
| Reference |
|
|
| Job Type |
Full-time |
|
| Job Status |
Sourcing |
|
| Date Posted |
Friday, 18 January 2008 |
|
| Location |
|
|
| Duration |
|
|
| Company Information |
INRIA
Laboratoire LIG, ENSIMAG
antenne de Montbonnot, ZIRST, 51, avenue Jean Kuntzmann
Grenoble,
38330
Website: http://www.inrialpes.fr
|
|
| Job Description |
Volunteer computing is a form of distributed computing in which the general public volunteers processing and storage resources to computing projects, such as SETI@home, ROSETTA@home, and EINSTEIN@home. Currently, about 1 million hosts contribute 1.1 PetaFLOPS of computing power for 50 scientific projects. The majority of volunteer computing projects have applications that consist of independent and compute-bound tasks. At the same time, many important scientific applications (for example, protein folding) consist of computations with dependencies structured as a DAG.
One main challenge in supporting DAG applications is the fluctuating availability of desktop resources (due to user load, a machine being powered off, and etc.), which can block forward progress of the computation. Our approach will be to use task replication across multiple resources as a means of masking failures.
In particular, this position has three main objectives. The first objective is to understand how resource failures effect the execution of a DAG applications, and how these failures can be masked using task replication. This will involve developing a probabilistic model of application execution as a function resource failure rates and the replication factor. Validation of this model will be conducted using simulation driven by our availability traces of real volunteer computing resources. Applying the model, the second objective will be to design and evaluate replication strategies for deciding which task to replicate, which resource to replicate the task on, and when during the computation to replicate. The third objective is to implement the best strategy in the open source volunteer computing software infrastructure called BOINC (Berkeley Open Infrastructure for Network Computing), which currently runs across about 1 million desktop PC's. Platforms available for testing and deployment purposes include DSLLab (a test bed with 40 nodes over the Internet), Grid'5000 (8 clusters with over 2500 nodes), and XtremLab (a joint BOINC project with over 10,000 desktops spread across the Internet).
See http://www.lri.fr/~dkondo/positions.html for more info. |
|
| Qualifications |
Experience with parallel, distributed, or Grid computing, and probability theory is a plus. |
|
| How to Apply |
Email Derrick Kondo at dkondo@imag.fr
and apply here:
http://www.talentsplace.com/syndication1/inria/ukpostdoc/details.html?id=PNGFK026203F3VBQB6G68LOE1&LOV5=4508&LOV6=4515&LG=EN&Resultsperpage=20&nPostingID=1919&nPostingTargetID=4911&option=52&sort=DESC&nDepartmentID=19 |
|
| Email Resume To |
dkondo@imag.fr |
|
| Contact Info |
Derrick Kondo
Contact Email: dkondo@imag.fr
|
|
|
|
| Go back to last page |
Home 
Mobility Portal 
