Cloud Killed Grid…Not if You Ask our Customers

It has been a week or two since Dave Rosenberg at the Register wrote an article that sounded the death knell for grid computing. In the article Dave points to the new Amazon EC2 Compute Cluster instance service as the cause for this conclusion.

It has taken a while to sit down and respond, but I think it is important to provide an alternative perspective based on our customer activity and feedback that addresses some of the points in Dave’s article.

We don’t believe Cloud is killing Grid, we see it as an evolution of the options for distributed computing infrastructure, Clusters, Grids, Clouds or whatever comes next. It doesn’t matter what you call the environment, what matters to customers are key things like:
  • Application performance
  • Application support
  • Total Cost of Ownership
  • Ease of management
  • Time to deploy/flexibility
  • Scalability

As Dave states there was a lot of hype and activity in the academic community about grid. Our perspective includes a very strong Global 2000 enterprise customer it is important to understand how those organizations have adopted and continue to adopt grid (and clusters, and clouds) and how this new offering might add value ( not kill) their grid environments.

  1. Grid has been widely adopted in enterprise environments for computationally intense workloads. The companies (across many vertical markets) that have deployed Grids, use them as mission critical infrastructure to support key business processes (design, engineering, analysis, etc) and they have been able to achieve cost effective scale for these environments based on the adoption of commodity infrastructure. They are able to operate these grids at cents per hour, while driving resource utilization >80%, 24x7. Our customers’ demand for resources is ever increasing, so they look to whatever distributed computing infrastructure they can tap into for additional capacity.

  2. The new offering from Amazon which provides higher performance compute instances in the cloud, now starts to enable customers to consider leveraging the cloud for some of their Grid applications. The decision a customer makes to use a cluster in the Amazon cloud, or operate their own environment will driven by the considerations listed above.

  3. The cloud offering from Amazon is a good start, but still requires customers to configure and manage their environments for their applications. This is where companies with expertise in Grid and HPC like Platform can help customers make best use of this new offering. Individual customers are going to make the decision on how to leverage the cloud offering to extend and add value to their mission critical grid environment, and like most decisions on outsourcing, the answer will be “when it makes sense economically.” We are likely to see a hybrid model for many years to come as customers make the choice that is right for them and their business.

So Amazon offerings rather than killing grid will make grids more dynamic and make it more apparent that the evolution (and relationships/learning’s) of clusters, grids, clouds is a natural one.


Post a Comment