The enterprise is fast approaching a critical juncture in the development of virtual and cloud architectures – the point at which public, private and hybrid infrastructure starts to meld into a unified data architecture.
But while the idea of a fully seamless environment in which applications and data can burst across internal and external infrastructure at will is still a little fuzzy, the fact remains that cloud computing will soon become the norm for the enterprise, rather than the exception. And that means the pressure is on to convert existing silo-based infrastructure into a private cloud.
Too bad we don’t have a magic pill to make it happen. Instead, enterprise executives are forced into a soup-to-nuts evaluation of existing infrastructure, followed by a detailed plan as to how they intend to convert it into something that is still largely undefined. Industry-wide, however, the knowledge base is steadily increasing, and a quick scan of what some early adopters have learned can make the cloud transition a lot easier for all.
According to Allstate’s Jim Ditmore, the enterprise needs to take four key steps to prepare legacy infrastructure for the cloud. First, the entire application stack – from hardware on up – needs to be standardized, with no more than 20 server images on tap. After that, a clear-cut catalog and ordering process must be implemented, followed by a fully automated provisioning and deprovisioning platform that enables server images to be created within a few minutes. And finally, undertake a cost and allocation assessment in order to tailor the system to real, rather than anticipated, user requirements. Firms like JP Morgan Chase and Fidelity are already seeing 50 percent IT cost reductions through private cloud deployment.
This doesn’t mean, however, that the private cloud must follow a rigid design pattern. After all, one of the major objectives is to increase infrastructure and architectural flexibility. As BlueCat Networks’ Branko Miskov pointed out to our sister publication Enterprise Networking Planet, even standardization around a single hypervisor like VMware is a mistake, considering other platforms like Hyper-V and KVM can often provide high performance at low cost. And since the cloud will invariably lead to complex data environments, extensive automation and self-service provisioning is a must.
Of course, building the cloud is only the first step. In order to optimize performance, you’ll have to incorporate a number of practices that take you out of the confines of your own infrastructure. For IT consultant Bill Kleyman, these include data migration and control, with special attention to data loss protection and granular visibility into external infrastructure, as well as broad WAN optimization capabilities that anticipate the wide area network behaving increasingly like a LAN as architectures become more distributed. Also, it helps to have a clear idea of how many users your new cloud will be servicing, as well as their workload and application requirements and the physical distances your cloud will encompass.
But probably the most important aspect of the cloud is the realization that it is not the ideal architecture for every data function. As Ipswitch Inc.’s Alessandro Porro notes, the cloud is not the best for proprietary applications. Home-grown graphics and multi-channel warehousing, for example, will probably require more hands-on management than they’re worth in the cloud, which should be designed specifically for streamlining and aligning enterprise resources. As well, applications that do not lend themselves to distribution across multiple VMs are probably best left to traditional infrastructure.
The transition to a fully functioning cloud environment will likely be a long one, but that doesn’t mean the enterprise has the luxury of delay. For many organizations, the process has already begun, which means anyone who hasn’t yet made a move toward the starting line is already behind the curve.
Don’t panic, but do take careful stock of where you are now and where you want to be in the cloud, if only to avoid committing yourself to an infrastructure that does not meet your needs in the long run.