vPivot

Scott Drummonds on Virtualization

Private Clouds, People Consolidation, and Chargeback

11 Comments »

A couple weeks ago EMC’s CIO, Sanjay Mirchandani, visited Singapore and presented EMC’s journey to the private cloud.  I sat in on one of his presentations and was absolutely amazed by his cogent argument for VMware.  There may be no better evangelist for VMware and its role in the journey to the private cloud.  Sanjay white-boarded one thought–a parenthetical discussion–that reformed my understanding of the value of virtualization and the private cloud.

Sanjay was talking about the allocation of resources to IT projects.  His whiteboard figure showed one stack where project resources were in silos dedicated to the lines of business (LOBs).  In this model resources are not commonly shared across multiple projects. Here is a version of the figure Sanjay drew on the whiteboard.

Hardware and people are dedicated to the needs of the LOBs. There is little or no sharing between projects.

Most of you are thinking it ludicrous that any company today would fully separate infrastructure and people this way. Surely storage and network sharing is common across a company, right? I can tell you that in more companies than you think there are still rigid lines of separation between different business units. The most common reason for this is, “our engineers require special hardware and support that corporate IT cannot provide so we give the engineers their own IT.”

EMC includes in its vision of the journey to the private cloud a step one that shows the consolidation of network and storage. Most companies today place most storage on a shared infrastructure. And when VMware is first deployed, servers are quickly consolidated to reduce redundancy and improve utilization. I see that most of VMware’s customers show the following silos and shared infrastructure.

In phase one of the VMware journey, IT production, customers are consolidating servers that run IT's applications. The application owners are still resisting consolidation and are given customized applications on dedicated hardware.

The beauty of virtualization is that not only can the physical resources be shared, as any VMware demonstration will prove, but the people that support the infrastructure can be shared, too.  This concept is already understood by VMware’s more mature customers, who have been telling VMware for years that virtualization can save more money in operational expenses than capital expenses.  These savings are coming after thinning the ranks of dedicated infrastructure specialists and refocusing them on higher value opportunities.

But after hardware has been consolidated and is shared across LOBs, how is it possible to do a similar consolidation with people, the most expensive part of the higher levels of the stack? Sanjay says the answer is through application rationalization. Applications must go into the private cloud and they must be shared by common resources. This means applications must be chosen to:

  • Fulfill the greatest number of requirements for the greatest number of LOBs.
  • Be supportable in a virtual environment.
  • Run on a small number of OS builds also selected for their supportability in the cloud.

Success in this space greatly reduces the aggregate size of application support teams and can reduce software licenses through the power of large purchase discounts.  The result is further consolidation.

At this phase a smaller number of OS images are supported and application versions and vendors are limited. But the support structure is greatly decreased as a result.

So, through VMware’s virtual machines, it is easier, cheaper, and faster to deploy infrastructure to the LOB.  But the system retains one bottleneck: the interface between the LOBs and the virtual infrastructure.  Speaking as a performance person, this interface is an obvious source of latency and inefficiency.  Any person at this interface will simply stand in the way of virtual machine deployment.  If there are too many people in this role there is waste.  If there are too few, the interface presents a bottleneck to throughput.  The solution is to allow the LOBs to self-service.

Enter the private cloud.  Building a private cloud where the LOBs can self-service removes the last bottleneck and eliminates more dedicated resources (people).  The end product of a private cloud deployment is shown here.

Everything is shared in a private cloud deployment: hardware and people. Sharing increases the utilization and efficiency of both.

In my opinion the most important part of the private cloud is one of the least talked about: a transparent system of showing LOBs the cost of their use of the infrastructure.  With a proper chargeback system, with a fully transparent model, LOBs can understand what portion of the hardware and people they are consuming.  And by exposing pricing, IT organizations around the world enter the free market and show their ability to compete with public cloud providers.

VMware vCenter Chargeback is as important to the cloud as any end-user portal.  It is with its correct usage that owners of a VMware private cloud will either blossom or wither.  My customer discussions of late are starting to convince me that proper chargeback is not just a step in the cloud, but a key requirement.

11 Responses

Well done article and great read on private cloud and chargeback.

  • Chargeback puts you into the old IT model of being a cost center, blech. Be a strategic partner with the business, guide them to better solutions and you’ll find that the checkbook opens and both you and the business profit. I’m not sure what it’s like at most businesses but from working for SMB’s up to my experience with my current emplyer (an S&P 500 company) the cost of software licenses and business process inefficiency both completely dominate both hardware and IT salary costs.

    • Setting up cost models is certainly not sexy. But it is essential for survival. Amazon has published EC2 prices on the internet (http://aws.amazon.com/ec2/#pricing). If IT organizations choose not to provide theirs, their internal customers will continue to bleed to EC2, as it happening now.

      Accurate cost models will necessarily be pretty complex. But that is why tools like Chargeback are essential to automate the models’ calculation and application.

  • […] a recent article Scott Drummonds talks about the necessity in transparency of costs associated with spending between […]

  • The question in my mind is how do customers get there? Gradually? Big bang? I’m seeing solutions deployed (like VDI) not because the business case makes sense, but because it’s a step that takes them to a more agile way of working. Is this how it will happen, via a “transformation project”?

  • My experience on this not only about charge back, as you will also face a lot of challenges to Understand the Real Cost of every pieces in your infrastructure before you can even define the unit cost into the charge back mechanism.

  • Excellent article. I haven’t look much at private/public clouds, but your article clearly defines the concept and the business problems it can solve.

  • […] Drummonds – Private Clouds, People Consolidation, and ChargebackThe beauty of virtualization is that not only can the physical resources be shared, as any VMware […]

  • […] what we are talking about here is a realistic chargeback system.  I know that many of you roll your eyes and shake your heads each time I talk about this. […]

  • […] 元記事:http://vpivot.com/2010/06/24/private-clouds-people-consolidation-and-chargeback/ […]