Ask any IT expert to tell you what the chief advantages of the cloud are and you’ll invariably hear two key words: scalability and elasticity.
Scalability is easy enough to understand. It simply means the ability to quickly ramp up additional resources, which in the cloud usually involves scale-out techniques across widely distributed architectures.
Elasticity, however, is a fuzzier term. Elasticity implies the ability to shift and pool resources across disparate infrastructure so that data needs and resource availability can be kept more in sync, avoiding the wasteful practice of over-provisioning. The problem is that there is no clear-cut delineation between what is and is not elastic, meaning that anyone with a hosted service or enterprise platform can use the term for just about anything they want.
The OpenStack community has been one of the biggest boosters of the elastic cloud. Companies like Cloudscaling are staking their claims on the ability to provide broad platform compatibility, providing features like on-demand computing, intelligent resource management and public/private integration designed to foster production-grade cloud infrastructure. Under the company’s definition, elasticity involves optimization of cloud resources to support dynamic services for mobile, Web, Big Data, PaaS and other applications. That means elastic clouds must not only be open, but flexible enough to support self-managing, highly scalable applications that will come to define the changing data environment.
Automation is also a key component of elastic infrastructure. rPath offers a specialized elastic version of its Enterprise Cloud Adoption Framework (ECAF), which the company describes as pooled and standardized infrastructure offerings that enterprises can deploy according to their size and data requirements. This is the best way to support a unified application development platform, in which a highly abstracted user interface allows IT to be made available as on-demand services. To accomplish this, ECAF provides a dual-platform development environment that encompasses both infrastructure and application layers, bringing a high degree of automation and agility to both the development and provisioning process.
Elasticity is also finding its way onto top-tier hardware platforms, which are tasked with accommodating both higher volumes and more complex transaction and processing requirements as enterprise infrastructure becomes more diverse. IBM’s newest Power7+ machines, for example, feature a new system called Elastic Capacity on Demand for Power Systems Pools designed to make them more appealing to cloud, analytics and sensitive workloads. The company describes the package as a utility-like distribution system that allocated virtual processor and memory resources according to application needs. The intent is to not only boost utilization rates and energy management, but to improve application and data availability through more effective resource allocation.
Meanwhile, Oracle is touting the newest version of its Exalogic Elastic Cloud, the X3-2, an integrated hardware/software platform featuring the latest EEC 2.0 software. The system is said to support a 60-fold increase in application deployment density through techniques like Single-Root I/O Virtualization (SR-IOV), as well as a near tripling of available RAM from 2.9 TB to 7.7 TB. As well, the platform provides a third more cores using new Sandy Bridge Xeons, packing as many as 480 in a single rack.
Clearly, the enterprise industry is trying to capitalize on the success of services like Amazon’s Elastic Compute Cloud (AC2) and the Google Compute Engine. Elasticity is so ingrained into the idea of the cloud that any service provider that doesn’t profess to have a high degree of it is quickly laughed out the door.
But in the absence of a clear-cut definition of elasticity compared to, say, simple load balancing or on-demand provisioning of virtual resources, the term will continue to serve more as a marketing tool than as an actual system capability.
There’s nothing unusual about that, mind you, as technology firms have been bandying terms like “flexibility” and “operational efficiency” for years without offering a hint as to exactly what they mean. It’s the same as adding the word “hearty” to a can of soup to make it more appealing.
The cloud may provide many advantages over existing data infrastructure, but for the moment, elasticity is in the eye of the beholder.
