HomeCernVM and VML - Virtualization for Scientific Computing
CernVM and VML - Virtualization for Scientific Computing
Date: Friday, May 22, 2009
Time:10:00am - 11:00am
Location/: LBNL Bldg. 50B, Room 2222
Speaker:
Yushu Yao
Physics Division
Lawrence Berkeley National Laboratory
Abstract:
ATLAS is a High Energy Physics project that has a large calibration
involving thousands of physicists and developers. ATLAS software
development and data analysis has been mostly based on remotely
logging into Linux clusters. The fast rise of virtualization
technology has the potential to change this model, turning every
laptop or desktop into an ATLAS analysis/development platform. In
addition, the application of virtualization technology could also
turn the current batch clusters into more flexible and easily
manageable computing clouds.
CernVM is a Virtual Software Appliance to run physics applications
from the LHC experiments at CERN. The virtual appliance provides a
complete, portable and easy to install and configure user
environment for developing and running LHC data analysis on any
end-user computer (laptop, desktop) and on the Grid independently of
operating system software and hardware platform (Linux, Windows,
MacOS). The aim is to facilitate the installation of the experiment
software on a user computer and minimize the number of platforms
(compiler-OS combinations) on which experiment software needs to be
supported and tested thus reducing the overall cost of software
maintenance for LHC. CernVM Operating System, based on rPath Linux,
fits into a compressed file smaller than 100 MB and represents a
common platform that can host software frameworks of all four LHC
experiments. The experiment software stack is brought into appliance
by means of CVMFS, a file system specifically designed for an
efficient and 'just in time' software distribution. In this model,
the client downloads only necessary binaries and libraries as they
get referenced for the first time. By doing that, the amount of
software that has to be downloaded in order to run the typical
experiment tasks in the Virtual Machine is reduced by an order for
magnitude.
In the context of the CernVM project we are developing a suite of
tools and CernVM plug-in extensions to promote the use of
virtualization for ATLAS analysis and software development. The
Virtual Machine Logbook (VML), in particular, is an application to
organize physicists' work on multiple projects, logging their
progress, and speeding up "context switches" from one project to
another. An important feature of VML is the ability to share with a
single "click" the status of a given project with other colleagues.
VML builds upon the save and restore capabilities of mainstream
virtualization software like VMware, and provides a
technology-independent client interface to them. A lot of emphasis
in the design and implementation has gone into optimizing the save
and restore process to makepractical to store many VML entries on a
typical laptop disk or to share a VML entry over the network.
In this seminar the speaker will first introduce the computing model
of ATLAS, then some ideas on CernVM and VML, and at last its
potential with cloud computing.
Host of Seminar:
Paolo Calafiura