Posts Tagged ‘in-memorgy data grids’

Elasticity

May 18, 2009

Elasticity – the property of a substance that enables it to change its length, volume, or shape in direct response to a force effecting such a change and to recover its original form upon the removal of the force.

This property is in essence the defining characteristic of a cloud; the ability for the cloud to grow and shrink based upon the resources required to support the services being delivered. Now to some extent much like most algorithms for random numbers which require a seed there is a pseudo aspect to cloud elasticity in that the cloud is always bounded in some way by the physical resources (read that as the physical machines that host the virtual machines) that are available to the cloud, so elasticity really reflects a behavior experienced as opposed to a true physical characteristic.

Probably the most obvious illustration of cloud elasticity experienced is with Infrastructure as a Service (IaaS), Amazon EC2 being the most famous example, where compute resources are made available as a service and the user of the service can scale up or down the amount of compute resource required in a frictionless manner.

One other example that I have been noodling is related to distributed caching platforms; high performance distributed in-memory object caches / grids, which by design run across multiple servers/machines. It’s not a leap of faith to see distributed caches running as a service in a cloud, with the cloud growing and shrinking to accommodate the size and requirements of the cache. I manage such an offering called Websphere eXtreme Scale and Billy Newport has been discussing some of these scenarios in his blog.

Another key aspect of elasticity that has me thinking is elasticity as it relates to business models, particularly pricing and licensing of software running in a cloud, here my thoughts center around the provider of a cloud service and the licensing of software that supports that service.  As we evolve to running more and more virtual servers in a cloud, whose role and function may vary by the minute, the notion of license pools will need to be fleshed out to support this use case.  One might envision something analogous to the cell phone model, with a license pool representing cell minutes, and overage structures being available to deal with license overruns. These kind of structures will drive new capabilities and indeed philosophies when it comes to billing and metering and the notions of software compliance. 

Through this year, both aspects of elasticity; delivery of middle ware services and licensing models, will have a position of prominence on my to-do list, who knew that high school physics lab on elasticity of objects (remember the one with the metal ball and ring) would prove so relevant!

Advertisements