At the National High Performance Computing and Communications Conference held this past week in Newport, R.I., (www.hpcc-usa.org) “Grid” was definitely not a four letter word.
This “insider's conference,” now moving into its 20th year, still maintains a certain elite status. Senior technology visionaries and policy influencers once again gathered on Newport's beautiful Goat Island to talk about initiatives, technologies, products, research and funding that represent the future of High Performance Computing. It should be no surprise that Grid was included in almost every presentation.
Grid was mentioned as part of HPC strategic directions for a number of private, public and government organizations, supported by some lively discussions of Grid being the enabler of convergence between HPC and enterprise computing.
Here's an interesting factoid: From an informal survey I conducted among attendees — keeping in mind this was an HPC conference — there were just as many attendees that read GRIDtoday as those referencing HPCwire (a sister publication of GRIDtoday, published by Tabor Communications Inc.). Grid is undoubtedly the hottest topic in HPC.
While presentations from Intel and SGI focused more on “thinking inside the chip” — the core technology advancements and ever-complex challenges of silicon fabrication — the majority of the presentations were focused on “thinking outside the box” — on application requirements with never-ending computing needs and the realization that organizations need to start thinking differently about how they build their computing infrastructures.
Several attendees I spoke with were surprised to see an obvious trend being unveiled: high-end computing applications across many industries are now relying on Grid infrastructures to meet their computing and communications challenges. It's happening today. And just witnessing the interaction during the breaks and networking events made another point quite obvious: the dance floor, as we say, is getting crowded.
With all these technology advancements, I find it refreshing that business still gets done by people dealing directly with people. One attendee told me he was able to make more personal introductions and hand out more business cards at this intimate gathering than at another event he recently attended with several thousand people. I concur. And, better yet, everyone wanted to talk about Grid.
I heard descriptions of Grids, clusters, networks, shared departmental systems — all being referred to as Grids. But more on that later.
If it Walks Like a Duck …
While working on this column, two odd-looking ducks landed on my balcony. I was staring at them trying to figure out what was so different about them. It made me think: if it walks and talks like a duck — it might just be a duck. But does it really matter? All these implementations might just be a Grid.
A number of analysts and industry observers, this writer included, have said for some time that the many, varied definitions of Grid, or lack of a consistently-used definition, has been a barrier to wider spread adoption. I no longer believe this is the case. If an organization believes that its computing infrastructure, network and shared resources constitute a Grid, then it serves their purpose, regardless of it meeting any textbook definition of a Grid.
This broad-based use of the term Grid might, in fact, be helping us to embrace the potential of what a Grid infrastructure can offer. That's not a bad thing.
I think one of the best known Grid visionaries and industry leaders, Dr. Wolfgang Gentzsch of MCNC (North Carolina) made this point very well during his panel presentation. His message: Don't get hung up on the definition of a Grid. All the efforts that companies today describe as Grid infrastructures are helping to move the industry in the right direction. I agree.
What is Really Driving the Momentum Toward Grid Infrastructures?
Walking among this group of HPC gurus, I was curious. Why has Grid become so important? Is it because the standards are finally evolving? Is it because of the overwhelming cost savings to be realized? Is it all about more efficient IT utilization? Not from what I heard.
As most agreed, the reason for this recent momentum comes with the realization from a number of these organizations that they need to start thinking of Grid a bit differently. “Grid is not just about saving IT dollars.” “Grid it is not just about more efficient utilization of resources.” The emphatic point made by a number of forward-thinking industry leaders is that Grid is an enabler of a new way of thinking, a new way of doing business and a way to do new things.
I think there are two things we can expect to see this year. New technology to support the computational challenges of unmanageable amounts of data, and new thinking on how to build computational infrastructures — the quest for a cyberinfrastructure — and the realization of Grid as being the foundation of high-end computing advancement.
I've met with a number of companies over the past year that were simply in a rut — trying to remain competitive by attempting to apply new technology without changing the way they thought about or approached the problems. You can't simply apply old problem solving to a new computing paradigm.
Grid could enable the convergence of HPC and mainstream computing. That's huge.
Grid is fostering a new way of thinking. Grid is even driving the investment, both in money and resources, into rethinking business processes and usage models.
Grids can be built to serve specific (work group, departmental) purposes and, certainly, the enterprise can benefit from organization-wide virtualization of resources. Some organizations have chosen to describe Grid as part of the bigger picture — part of the Service-Oriented Enterprise — but my thinking leans toward Grid actually being the bigger picture. That is, Grid is the infrastructure of the new enterprise architecture. Grid is the architecture of the future.
A Grid implementation offers the potential of unlimited access to data, with no dependency on location, and not just to access or retrieve the data, but also the ability to manipulate the data to accomplish things previously impossible. It's about virtualization, and that has very broad appeal well beyond the HPC community.
As we've learned historically, new technology forces us to transform our business processes. Adopting new business processes often allows us to tackle new business problems or challenges. The enterprise of the future will be a mesh (or a Grid if you will) of all resources available to take on a given problem at a specific point in time. Dynamic, virtual, scalable, seamless, transparent.
The concepts of utility computing or autonomic computing — and even the Service-Oriented Enterprise, the architecture and infrastructure discussions around the enterprise or the architecture of the future — make much more sense if a Grid infrastructure is already in place. For the most part, we are all talking about the same thing.
The adoption challenges for implementing a Grid infrastructure fall into three categories: technical, cultural and administrative. We plan to explore these challenges in-depth with a significant state-of-the-industry market analysis report to be released in the fall of this year.
But for now, here's the latest buzz: Grids are far from mature, and they bring along their own set of challenges, with the political challenges not to be taken lightly. However, the effort is worth the pain — don't get hung up on the small stuff.
About Mike Bernhardt
Mike Bernhardt is CEO of Grid Strategies Inc., a niche market analyst, consulting and advisory group offering guidance, counsel, and business and market planning services for clients of all sizes working in and around the emerging Grid industry. Learn more about the company at www.GridStrategies.com.