March 25th, 2009 11:46 am
Posted by Gary Tyreman
Tags: Dell, Ganglia, Globus, Grid Engine, HPC, ICR, Intel Cluster Ready, OFED, Open Source, UniCluster
HPC is a big word with many meanings to many people. I prefer to refer to it as a bucket of capabilities (someone recently referred to it as a way of life) or components that are used to solve extraordinary problems. HPC in and of itself is a large, complex set of components. Consider the history in the context of the theme of openness and standards.
HPC is complex. I haven’t forgot that – but frankly I dismiss that as a crutch. Personally I think there are harder things like learning when to keep your mouth closed in front of your wife so as a to not create more work for yourself (like my stone work, deck and gardening that keeps me busy when the Canadian winter melts into Spring until it returns in the Fall….).
HPC has seen its fair share of openness with many amazing open source projects like Globus, Grid Engine, OFED, Ganglia, Nagios, Lustre to name but only a few. Many of these projects are represented on www.grid.org or included as a component in UniCluster.
HPC has also been the epicenter of commoditization – from engineered performance, proprietary and expensive capability computing machines to off-the-shelf, bundled commodity computing systems available as close to turnkey as possible.
We’ve also had some success, albeit limited, in standardization through the work of many organizations like Open Grid Forum: DRMAA, HPC Basic Profile, BES are some recent specifications.
HPC has taken a BIG step forward through the introduction of Intel Cluster Ready, a specification and program to make it easier to run applications off the shelf. Intel® Cluster Ready is a great example of making complex HPC systems deterministic. In 2005 I set up a meeting with Intel and Dell and we talked about the challenges facing the broad adoption of HPC in savvy and non-savvy commercial organizations. Complexity, engineered performance and commoditization were on a crash course and it was slowing the market, particularly with entry-level customers. We agreed to work together to resolve ISV enablement. The outcome of this ‘French Riviera Summit’ (I was in the South of France during the conference call) was the Intel Cluster Ready program nearly two years later.
A key enabler of this program was open source software. That is, components and reference implementations were readily available because they were open and adoptable. The program would have been hard pressed (and less successful) if elements were closed and unavailable.
JOIN THE CONVERSATION
You must be a Registered Member in order to comment on Cluster Connection posts.
Members enjoy the ability to take an active role in the conversations that are shaping the HPC community. Members can participate in forum discussions and post comments to a wide range of HPC-related topics. Share your challenges, insights and ideas right now.
Login Register Now