 |
|
DAILY NEWS AND INFORMATION FOR THE GLOBAL GRID COMMUNITY /
|
Special Features:
CCT HPC MANAGER ON THE OUTLOOK FOR
LINUX CLUSTERS
by Derrick Harris, Editor
GRIDtoday interviews Brian Ropers-Huilman, CCT's HPC manager, about the
current state and future of Linux clusters and their place in HPC.
GRIDtoday: What are the factors that currently contribute to growth of the
Linux HPC market, and in particular what are the major inhibitors that
challenge that growth, both technically and economically?
Brian Ropers-Huilman: The largest factor helping the current growth of the
Linux HPC market is the simple fact that both the hardware and software
systems have evolved to a point where it is trivial for any organization to
assemble at least a modest cluster. That being said, the major inhibitors are
that, while the software stack has matured, there are still far too many
fragments that have not been bound together sufficiently. The cost of free
software is the price paid for very experienced and high-level cluster
administrators.
Web services, utility computing, .NET, CPU harvesting and distributed
computing are just a few of the technologies that fall under the Grid
computing umbrella. Gt04 -- a premiere enterprise Grid computing conference
targeting industrial and commercial users -- will gather experts, and outline
strategies and road maps for Grid deployment. For more information, visit
www.gt04.com.
Grid computing is here!
GRIDtoday: To what extent is current Linux cluster deployment influencing
future growth?
BRH: Current Linux clusters have proven that the concept is technically
feasible and sound. Computational clusters are relatively easy to do. However,
what is emerging as a result is the fact that there are few tools available,
to ride on top of the clusters, to handle the data that these clusters either
generate or need to act on. More organizations are able to deploy clusters to
serve their needs, but the toolsets for managing the ever-growing datasets
does not exist in mature form.
GRIDtoday: What changes are in store for Linux HPC clusters in the next 5
years?
BRH: I predict that Linux clusters, in the main-stream, will still be the mid-
sized systems we see today, on the order of 256-512 node systems. These
clusters will have very large memories and will be attached to massive amounts
of disk. Toolsets will be developed allowing cluster integrators to highly
tune performance on the clusters in a manner much easier than what exists
today. While the cluster itself has been the emerging trend over the past five
years, the attached storage and interconnecting networks will be the focus for
the next five.
GRIDtoday: Who will be the main users 5 years from now?
BRH: Clusters, given their commodity nature, will find their way into more and
more business organizations and may even become as common as desktops on all
employees desks. As such, I see the main Linux cluster user 5 years from now
being banks, financial institutions, security organizations, petro-chemical
and pharmacological companies, and, of course, the traditional university or
researcher.
GRIDtoday: Are users leading the path or are they merely adapting to the
technology path that is presented to them?
BRH: Users do play a role in the evolution of Linux clusters. There is a
feedback cycle where the researchers develop new technologies, systems
integrators incorporate that technology, cluster administrators deploy that
technology, and then users attempt to utilize that technology. Based upon user
feedback to the administrators, they will push ideas to the integrators and
the researchers. The technology is changed and the cycle starts anew. For the
most part, however, I see users as following the technology rather than
leading it. I still see universities and research organizations as the prime
movers in Linux clusters, but I think we will see more and more industry
involvement as well.
GRIDtoday: Has there been any real impact of open source to the growth of the
Linux HPC market, and if so please cite specific examples?
BRH: There has been a definite impact of open source on the growth of the HPC
market. Obviously the Linux kernel project itself is a readily identifiable
open source project that enables the Linux HPC market, but there are more.
Projects such as MPICH, out of Argonne National Labs in collaboration with
Mississippi State University, Rocks Linux, part of the National Partnership
for Advanced Computational Infrastructure (NPACI) project out of the San Diego
Supercomputer Center (SDSC), and the Open Source Cluster Application Resources
(OSCAR) project, out of Oak Ridge National Labs are prime examples of how open
source software has helped the growth of the Linux HPC market. These packages
are often the core components.
GRIDtoday: Is there any future to "value-add" Linux systems, or will economics
eventually squash all attempts to increase vendor profit margins?
BRH: I do see a future in value-added Linux systems integrators. This added
value will help drive adoption by more and more organizations. Value such as
custom scripts to manage cluster power, user creation, log management, and the
like all contribute value to a cluster. Vendors who develop a full suite of
such helper applications will be able to penetrate into the non-traditional,
non- research markets.
GRIDtoday: How do businesses plan to cope with this pressure to reduce cost by
eliminating these value-add proposals?
BRH: I do not think that Linux cluster businesses are feeling any pressure to
reduce costs by eliminating value-added proposals. On the contrary, the Linux
cluster integrator who is able to distinguish themselves with such value-added
tools will emerge as a market leader. I believe early Linux cluster vendors
were simply too far ahead of their time, but truly feel that Linux clustering
technology, as well as many other Linux-based solutions, is poised to explode
into the market, reaching those organizations that would not have considered a
Linux solution previously. Cluster vendors who add value through utilities or
outstanding service will emerge as market leaders.
|