iaas

Things are heating up on the private IaaS front.  As of late, the hot question of us has been “What do we choose to build our private cloud?”  Naturally, our response is, It depends.

In determining a direction forward the background of the technology development for each is critical.  For all intents and purposes VMware is a virtualization platform.  It was originally developed to solve pain points related to IT operations, namely increasing resource utilization through consolidation.  It does this very well having long ago garnering the title of market leader.  Administrators are very familiar with ESXi, vSphere, vCenter and the cast of supporting tools.  Extending the capabilites as VMware has matured (both as a company and product), additional functionality aligned with operations:  DR, charge/showbacks, etc., has been added.  The mature set of products lends itself extremely well to traditional application workloads.  Think Exchange, Oracle, and other monolithic applications.  Lydia Leong at Gartner calls this “cloud-out.”  And I agree – originally built for IT challenges now being developed and sold as private IaaS.

Let’s now consider OpenStack.  I’ve heard OpenStack referred to as “more of a project than product”, I believe this to be true.  Conceived and built from the ground up by Chris Kemp and a group of talented folks at NASA it was aimed at solving issues related to leveraging cloud computing:  Quick spin up of instances (VMs in VMware parlance), including networking and storage, under a self service provisioning model.  These environments are aligned with the ephemeral, distributed nature of cloud-native applications and research.  Unfortunately there were other items required for IT (charge/showbacks and multi-tenant billing), and these were added later.  OpenStack and its distributions can therefore be called “cloud-in.”  While we can certainly run traditional workloads on OpenStack, it is not really suited for this.  For example, there isn’t a concept of HA clustering.  Why?  Because the applications must be designed from the ground up taking into consideration that applications and components will fail, and that the data will eventually be the same across the application stack.  CAP theorem in practice.

Ok, so now we know more about VMware and OpenStack and their origins…but, does that really help us choose?  Not quite yet.  We also need to evaluate the business.  Only through this lens will customers achieve the desired results.  One should consider:

  • Is there a deployed virtualization platform today?
  • How mature is the virtualized environment?
  • How large is the organization, and is it poised for growth?
  • What is the experience level with Linux on the SA/IT team?
  • Is there a preferred hardware vendor?
  • Does the organization utilize public cloud computing today, and if so, how and with whom?
  • Is there a significant push to reduce capital expenditure?
  • Is end-user/LOB self provisioning a goal?
  • How critical is multi-tenacy (with departmental charge/showbacks)?
  • Got Windows?

I wish that I could insert a hip infographic now, with which percentage of each client with what amount of technology and different business challenges chooses based on the above.  But it isn’t that easy.  Sure, we se a ton of VMware, and rightly so.  It is a heck of a product and when considering other items (like public IaaS) and it makes good sense.   On the other hand, having an Amazon EC2/S3-like private IaaS cloud is interesting and fits a ton of business needs like a glove – at a potentially far lower cost.  Cool, right?

Summing it all up…the technology direction just doesn’t require a look at the technology itself, but a detailed assessment of the business and requirements.  Depending on your shop there may not be a clear “winner”, in fact the right approach may be implementing and operating both VMware and OpenStack.

Tags: , , ,