Is grid relevant? Modern grid computing and today’s enterprise.

150 150 CloudPulse Strategies

Grid computing baggageFor many, the phrase "grid" continues to carry the baggage of first generation grid computing, which was marked by complexity, rigidity and worst of all a lack of relevance for the vast majority of problems faced by mainstream businesses. 

This is certainly no longer the case with grid-based application platforms like Appistry EAF, which are easier to use, flexible and built to support service-oriented and transactional applications–the bread and butter of modern IT–in addition to those with a CPU-intensive nature.

While trends toward cloud computing and the ever-increasing need for scalability are prompting more and more businesses to turn to modern grid technologies, I still run into people who associate grid with SETI@home. That’s why David Abramowski’s article, Is Modern Grid Computing Applicable to Business?, resonated so strongly with me. The article, posted Friday to the Web 2.0 Journal site, answers its own question with a hearty YES!

"It’s no longer relevant to define grid computing in terms of expansive pioneering volunteer projects such as SETI or Folding@home. It’s also unnecessary for commercial applications of grid computing to be defined solely by consumer cloud computing success in the Web 2.0 sphere. By dedicating networked servers to application hosting, grid computing enables scalability and efficient application deployment to the enterprise, accelerating ROI for everyday businesses."

David’s article focuses on five key areas which he feels provide the most compelling reasons why grid represents the future of business computing: scalability, speed, deployment, pervasiveness and cost.

In addition to being on-target, his list is well ordered. In particular, whereas many first-generation grid buying decisions were based upon cost savings from increasing the utilization of existing "big iron," businesses today are adopting modern grid technologies because they enable them to achieve things that were previously impossible or uneconomical at a given scale.

Abramowski recognizes this implicitly, citing scalability as "the very essence of distributed or grid computing," and noting that:

"As physical enterprise IT systems increasingly adopt more decentralized infrastructure, distributed computing will unlock new business models, improve business competitiveness, and facilitate rapid change and structural migration."

The one driver I’d add to David’s list is reliability. In a world where applications are hosted in clouds of inexpensive commodity servers, applications must be able to survive the inevitable node or VM failures. The inherent redundancy of the grid environment makes a high degree of reliability possible… if it’s provided for by the grid platform. This has ramifications for disaster recovery and COOP (i.e. continuity of operations) as well.

David’s final sentence poses a precondition for a happily-ever-after scenario: "If applications play nicely with grids…" In the past, that would have been quite a mouthful. Traditional grids and distributed computing technologies were hard to use, and pushed all the responsibility for and complexity of leveraging the grid onto the application’s developers.

With modern grid-based platforms, however, that’s not a tall order at all. Much of the complexity of building grid-enabled applications has been eliminated. Developers can now focus on their apps and not the environment in which they’ll run. And even existing applications that know nothing of the grid can be made to play nice without modification.

Modern grid approaches are a great thing for grid, and a great–and highly relevant–thing for today’s businesses.

  • Rowland

    Don’t forget that the scalability issue in distributed computing has almost *always* been about making algorithms or similar computationally intensive codes execute faster. You’d write your codes that could either parallelise easily or pass onto a decent resource manager to farm out to a super computer or commodity cluster.

    Grid computing typically involves the sharing of organisational resources with other like-minded organisations – consider the US national labs that use Globus, or the ever-increasing EGEE grid hosted by CERN and other academic institutions across the EU.

    One of the problems businesses have with Grid computing are the security aspects, protection of IPR. At the same time, they want to reap the benefits of computation resources they would otherwise not have access to. Academia has not had so much of a problem since computation resources are purchased on public funds and would not eat into profits – not true for business who have to look at the bottom line.

    Reliability is an interesting issue, but like you said needs to be supported by the middleware *and* resource manager in question.

  • sam

    I’d certainly agree that the “Traditional Grid” approaches that have been developed to meet the needs of academia and big research have not stood up to the needs of mainstream businesses very well. A new type of grid platform is required.

    Your point about academia and profits is a good one also. We used to joke that for the academics, software is expensive and people are cheap (i.e. grad students), but for businesses the opposite is the case. Funny but true!

Leave a Reply

Your email address will not be published.