CERN openlab II - Platform Competence Centre
Virtualization can allow computer resources to be
used more dynamically. Virtual machines have traditionally been used
in mainframes to provide separated execution environments to
different users. A recent renaissance in virtualization has been
triggered by increasing computing capacity of commodity hardware and
changes in usage patterns, and novel applications for virtualization
have emerged. Xen and VMware deliver virtualized platforms on
commodity hardware, with computing efficiency close to native
performance. Future computer architectures from Intel and AMD sport
extensions for making virtualization more efficient.
The GRID can benefit from virtualization
in several ways; a few are listed in the following:
Multiple guest OSs can run
simultaneously on a single host. This allows different
distributions or versions of distributions to be
provided simultaneously. This further allows for
flexibility in choices of libraries and toolkits, and
distribution upgrades may be performed without
necessarily replacing the existing distribution.
Rather than launching a computational
job, the user can launch a complete guest OS
distribution. The responsibility of the complete set of
software to be run on a node can be left to the user.
Software which is run inside a VM can
not negatively affect the execution of another VM. A VM
is an isolated execution environment, which gives the
user the illusion of running on a separate physical
machine. Malicious or faulty code may not affect
anything except the guest OS running inside the VM.