Different Product Approaches to Cloud Infrastructure Software

By Jay Judkowitz (Profile)
Share
Friday, August 19th 2011
Advanced

This is the last in a series of four articles discussing infrastructure as a service (IaaS) clouds. The articles started with basic level setting and here we conclude with an industry assessment. The topics for the series are:

  1. Cloud 101
    • What is cloud
    • What value should cloud provide
    • Public, private, and hybrid cloud
    • Starting on a cloud project
  2. Application taxonomy, what belongs in the cloud, and why
  3. What you should look for in cloud infrastructure software
  4. Evaluating different approaches to cloud infrastructure software

Different Product Approaches to Cloud Infrastructure Software

In trying to meet these cloud requirements, different vendors with different product sets approach the problem in different ways. Each method has its pros and cons. As this is not a competitive analysis article, the comments here will be of a general nature and will not all be true of all instances of a given software type. Please consider this more as a way to seed your thinking than as a deep analysis of specific products.

Evolving a Virtualization Product Line

Using an existing server virtualization product line has the obvious and compelling advantage of starting with a built-in, mature, widely adopted, and well understood technology that is an absolutely critical basis for cloud infrastructure.

That said, there are drawbacks to this approach. Virtualization started as a partitioning and consolidation solution. Later, distributed management tools were developed that were targeted at IT staff and that were not at all appropriate for end user consumption. Adding additional layers of abstraction and workflow into the management stack to get to a self-service, multi-tenant cloud is potentially very complicated and hard to manage end to end.

Also, server virtualization solutions have historically catered to the sorts of applications that were most widely deployed in corporate datacenters and that could benefit the most from pure partitioning and the value-added services that could be wrapped around opaque VM containers – monolithic, stateful client/server apps running in production.  But, as we discussed in the second installment of this series, these are among the last applications that should be sent to the cloud, not the first.

Our recommendation for existing virtualization management products operating legacy workloads is to continue using them where they are the best fit and where the cost is justified. There is no clear benefit to migrating away from them, especially in the near-term. However, when looking at new projects with applications that are great fits for cloud, one should consider a pure cloud play today. But, make sure to set an expectation that these pure cloud play vendors ultimately evolve to either directly manage or at least federate to existing virtualization solutions, so that ultimately you get a common end user interface for your IT deployments.

Evolving a Systems Management Product Line

Systems management vendors have some great things to bring to the table. Like server virtualization solutions, they are mature, well understood and widely adopted products. In addition, they have historically focused on solving the problem of heterogeneity, which is critical for cloud. Lastly, they share the cloud mindsets on high scale and tightly controlled workflows.

However, there are issues here as well. Systems management tools, like server virtualization product lines, have historically been targeted at IT staff, not end users. Also, the basis of the workflows is generally tied to ticketing systems – formal requests and approvals. While this sort of workflow is fantastic for compliance purposes and is very IT Infrastructure Library (ITIL) compliant, it is too slow and has too much operational overhead for competing against a true cloud for workloads that can’t abide that level of manual process. Provisioning needs to be much more dynamic and self-service, with constraints placed ahead of time and evaluated by the system, rather than reactively being applied by a human. While a good cloud provides a very controlled and regulated delegation of permissions with great audit trails and other compliance functionality, cloud’s elastic and self-service nature fundamentally put it in opposition to ITIL-type processes.