Many people who have seen me present over the past year have heard me discuss the notion of complexity and who truly benefits from it. Given the current state of IT budgets and resources, the time is ripe to take a closer look at this issue. Most organizations that I work with are grappling with mandates to be more efficient and responsive to business or department needs (said in another way, more agile) and to improve application availability, all while maintaining a flat budget. These mandates often lead to public, private and hybrid cloud initiatives that include an emphasis on high degrees of automation. What is the Goal? A typical first step on the private/hybrid cloud journey is to look at the innovators in the public cloud for inspiration and then either adopt those solutions and/or work to build an internal private or hybrid cloud. This is where the problem often starts. Look at the major public cloud IaaS providers and you will notice that they all share the same common architectural tenets: A single x86 hypervisor platform and common virtual infrastructure layer A single purpose-built cloud management platform A stack whose value is mostly derived from software and not hardware Now consider how those principles compare to some of the private cloud architectural proposals you’ve probably seen. Over the past two years, I have seen dozens of private cloud architectures, and I place most in the Frankencloud category. Frankenclouds are often built from a collection of parts using a somewhat democratic process and many either fail completely or come at too high a cost. Let me explain. From a political perspective, it’s far easier to allow each IT group to keep working with their favorite technologies than to standardize on one set of integrated components. Multisourcing is often encouraged as a means to potentially save on acquisition costs. So while a private cloud project may begin with a goal of emulating successful public cloud IaaS implementations, the resulting architecture may look nothing like it. Common attributes can include: A multi-hypervisor stack with separate management silos One or more cloud management platforms that orchestrate across a variety of virtualization, storage, networking and security components A stack that has both hardware and software optimized components If multisourcing throughout the cloud IaaS stack is such a good thing, then why isn’t it pervasive in the public cloud? The reason is simple. It’s not. That said, enterprises are often encouraged to multisource virtualization, storage, networking and compute infrastructures, among other layers. The reason why: Complexity is great for profits! Many traditional vendor and consulting practices have business models that depend on high degrees of complexity and the professional services...

Read More