The Gartner report on Cloud Computing identified five trends that would impact cloud strategies in the Enterprise. As EGI finalises its strategy for supporting research communities how do our plans relate to their trends?

• Formal Decision Frameworks Facilitate Cloud Investment Optimization

Optimising investment is not just about capex vs. opex, but its about deciding where does an organisation add unique value to the services they are delivering, and where they do not to consider out-sourcing the delivery of those services. Within EGI we are able to tune the hardware within our resource centres to the specific needs of particular research communities to match their workloads. For workloads that can operate effectively on the infrastructure provided by public clouds, should we not see how we can offer such resources alongside others in EGI and thereby leverage the benefits from their economies of scale to our research communities?

• Hybrid Cloud Computing Is an Imperative

EGI has always been about offering a uniform service interface across resources in different organisations - a hybrid infrastructure. As we start exploring how to provide a uniform interface between a researcher's local institutional private cloud, to the private clouds in your collaborator's institution, to public clouds provided commercially, we will inevitably have a hybrid infrastructure. This is where EGI's experience in building grids of computing clusters can be used to builds grids of clouds. The work that has been undertaken by the Federated Cloud Task Force has already helped identify some of the issues in building hybrid clouds and will be explored in more detail over the coming months.

• Cloud Brokerage Will Facilitate Cloud Consumption

Although EGI does not charge for resources, it effectively provides a brokering service between the research communities it supports and its resource providers. A similar service will be needed as we federate private clouds coming from the NGIs where the brokering will initially be driven by capability, availability and access. But once commercial resources are offered alongside public sector resource, brokering for cost and cost-effectiveness add another dimension to the problem. Having a better understanding of the cost of our own resources is an essential first step in this direction.

• Cloud-Centric Design Becomes a Necessity

For many of our current research communities and their applications this is less of a concern. They have already designed their applications to deal with elastic resources - be it different numbers of available worker nodes in a batch cluster in a 'push' model of computing, or varying numbers of pilot jobs in a 'pull' model of computing. Regardless as to if the computing resources are on physical or virtual hardware the ability to cope with dynamic resource levels is already established.

• Cloud Computing Influences Future Data Center and Operational Models

As commercial cloud computing matures will EGI continue with its focus on running hardware within its resource centres? As the diversity of cloud computing offers increases (see the recent Amazon HPC instances) will our resource centres continue to remain cost competitive? This is certainly one aspect to be considered alongside data privacy, trust and other issues.

Summary

So, how well aligned is EGI to the trends taking place in Cloud Computing? Many of the same issues facing the Enterprise in the adoption of Cloud Computing are reflected in the discussions taking place in EGI as we consider where we provide value to the research communities we support. Is it the hardware or the expertise on effectively using the various software solutions to support their research? Is it cost effective to have 350 resource centres or should we concentrate our efforts in a smaller number of resource centres? Should we move to a more commercial model where individual public-sector resource centres compete alongside commercial public clouds to provide resources to research communities?

What do you think? 

Responses[1]

Posted by Steven Newhouse on April 7, 2012 9:25:00 AM CEST #