Tuesday 17 August 2010

Cloud Models

The hype surrounding cloud computing is tempered by concerns over data security and persistent storage costs.  The elasticity in capacity provided by the cloud is not the most efficient way of provisioning long and medium-term data storage - just as renting a car by the day isn't the most efficient way of owning a car.  Security issues for many organisations render the public cloud inaccessible.

Organisations with concerns such as these are turning to private and hybrid cloud delivery models to take advantage of the rapid deployment, self-service models of the cloud - but leverage their shared service infrastructure to provide IT access internally. 

The search for the ideal cloud deployment model is based a very specific efficiency formula.  Which is the most efficient way of delivering flexible IT capacity, in the most risk-free way? 

Over time, the multi-tier architecture delivered in today's datacentres may come to incorporate public cloud components, yet leverage internal data stores.  These hybrid cloud models appear to offer the best benefit, yet security concerns still linger wherever sensitive data is exposed beyond the firewall.

Private cloud models are simply an extension of the provision of virtualized capacity on request.  With service-based billing, and an assumption of risk by the infrastructure provider - private clouds can offer benefit in a shared services environment.  Data is held and transmitted securely, lines of business will pay only for the capacity they consume, and provisioning will be made quickly and by the end user.

Yet do these models truly offer the cost/efficiency of a well-run data centre?  Leveraging the cloud means taking advantage of the compute power available throughout the network.  As an extension of grid computing - this truly does make sense for efficiency.  Yet still there is a lingering, niggling doubt.  Renting cars in the long term rarely represents an efficient way of ownership - is IT capacity really different??

No comments: