A note to applicants: given my volume of email, I unfortunately cannot respond to all applicants. I do read all applications however, and there are a number steps that you can take to make your application stand out. My main suggestion is to read the position description and explain how your background or interests suits this positions specifically. Give specific details about how you can contribute, or, even better, suggest new ideas or comments. Simply stating that you are interested, repeating your resume, or your rank X in course Y makes your application difficult to evaluate in the context of these positions. Applications where it appears that you are spamming a list of open positions will be ignored. Anyhow, good luck!
Internship subjects: Positions are now closed.
Proposed Postdoc Projects
ADVISOR: Derrick Kondo
TEAM: INRIA MESCAL
CONTACT:
dkondoATimagDOTfr
Laboratoire LIG
ENSIMAG - antenne de Montbonnot,
ZIRST, 51, avenue Jean Kuntzmann,
38330 MONTBONNOT SAINT MARTIN, France
THEME: NUM
TITLE: Why does my desktop fail?
DEADLINE: March 22, 2009.
DESCRIPTION:
With the increase in software and hardware complexity,
frequent failures of the desktop PC have become the norm
rather than the exception. Causes of failures can appear
anywhere in the hardware (for example, overheating CPU,
memory error) or software stack (OS, DLL's). However, the
symptoms and causes of failures in general-purpose PC's is
not well-characterized nor perfectly understood.
Hence, our goals are 1) to characterize failures of machines
connected on residential broadband networks and 2) to
identify potential causes.
Our approach for the first goal is to use the BOINC
(http://boinc.berkeley.edu/) middleware for monitoring and
recording the time and state of the machine immediately
before failures. BOINC is a middleware for volunteer
computing that runs across over 1 million machines on the
Internet. The result of our measurements via BOINC will be a
large-scale failure archive unsurpassed in breadth and
depth. Then we will use these measurements to characterize
how failures occur in time and space. For example, we would
like to elucidate trends, stationarity, periodicity and
autocorrelation of failures for hardware and software
components on a single machine, and also correlation and
independence for components across machines.
Then we would like to elucidate the causes of failures to
help explain the "symptoms" observed through the previous
characterization. Our approach for this goal will be to
apply statistical techniques to mine our measurements, and
to identify correlation and causality between process,
application, or hardware state and failures. The results of
this research will be essential for understanding the
reliability of hardware and software components of desktop
PC's, for reducing the negative impact of their failure, and
for developing proactive methods to avoid them.
REQUIREMENTS:
Experience with (Windows) operating systems,
and distributed, parallel, or Grid computing.
Background in machine learning, or statistics is a plus.