Tuesday, August 20, 2013

Showing Business Value with a Private Cloud





Whether people love it or hate it, there is a distinct value to moving toward a private cloud. Even if you only adopt certain aspects, cost transparency is a common-sense approach that helps business leaders understand their investments in infrastructure technology and helps IT professionals become more strategic partners to the business. Comparatively, the traditional role of IT just executing tasks for the business and not being able to show their underlying value to the company is a risky proposition as public cloud adoption begins to mature. It is past time that IT professionals embrace their role as service providers for their organizations.

In the past, companies looked to reduce internal expenditures by turning to co-location, outsourcing, and hosting providers. This was driven by the pains of not understand the benefit of technology and trying to reduce costs; owning and operating a data center is an expensive proposition. It requires spending capital on data center superstructure, IT infrastructure, and personnel to manage operations.

I want to share an approach that is simple, uses your current investment in your VMware infrastructure, and can provide a tremendous amount of value to your organization.

Here are some of the key aspects that make up a Private Cloud:
  • On-demand
  • Self Service
  • Pooled resources
  • Elastic
  • Measurable
  • Standardized
  • Automated
  • Resilient

I am going to focus on a few key aspects; mostly the items that help provide cost transparency, define service level expectations (SLEs), and how to incorporate them into your elastic and resilient vSphere environment.

First, let's start with defining service level expectations (SLEs). These are the standard, and non-standard, services that you want to provide your organization; and the value you want to place on those services. The grid below shows the traditional Gold, Silver, and Bronze SLE deployment. There are infinite possibilities for service class solutions when working at the cluster level, but make sure to design within the framework of your infrastructure budget. The gold offering would include the latest model servers, tier 1 storage, and N+2 cluster redundancy. Silver and Bronze would use older equipment, tier 2 and tier 3 storage, and N+1 cluster redundancy.

Additional service bundles can include consumption models, resource guarantees, restart order, management offerings, and packages. This gives your IT business partners a broad offering of service levels to meet their business needs.


Isolation in the contemporary model of application life-cycle for production, acceptance, and development worked because they were all on a single piece of equipment. However, in the past few years IT has learned through the birth of the cloud, which is nothing more than a highly virtualized and automated compute cluster, that we are capable of becoming much more service oriented and we can offer a portfolio of infrastructure service options. This enables our business partners to decide what service levels their applications require and removes it from infrastructure operations decision making.

Cost transparency, it is vital to make certain that your business partners know that you are not fabricating the show-back charges, below is an example of disclosing the in-depth cost breakdown for a Gold Level Virtual Machines. It is clear, concise, and helps build infrastructure cost awareness that should be assimilated into the application lifecycle.


I used density ratios to calculate costs, this was based on my experience for the average amount of virtual machines I could expect on a certain type of hardware; however you could use the rule of thumb of 4 to 6 single vCPU virtual machines per core to create your density ratio.

My density ratios were based on a four physical processor server with 8 cores that were hyper-threaded for 64 processors per host - scale up model. It also included 256 GB of memory, which easily could have been pushed to 512 GB of memory to balance out utilization more effectively.

I used a 60 virtual machine density ratio for small (1 to 2 vCPU VMs), a 30 virtual machine density ratio for medium (2 to 4 vCPU VMs), and a 15 virtual machine density ratio for large (6+).

The annual prices for the virtual machines should be driven from the size of the virtual machine and the services being provided by your IT organization. You will notice, we are offering a Gold Virtual Machine Custom which has a flat rate of $4,500.00 a year. This is to encourage the resource consumers to utilize a standard offering.


One of the individuals in my career that I consider a great mentor was from IT Finance. She told me, "You want to price your services to support the behaviors you expect." My target resource goal was for 10% Gold, 60% Silver, and 30% Bronze. When building out your service cost model, the IT Finance team is going to become your best friend!

Consider this, if you went to the local Ford dealership and you have the choice between a new Ford Focus and a new Ford Shelby Cobra at the same price, most people are going to choose the Shelby Cobra. You are going to run into the same mind-set with infrastructure resources. If you don't effectively cost the infrastructure services you are providing your business partners then they will certainly choose the best SLE available.



At the same time, you can further support your target resource goals by providing a governance heat map which defines the service level offering for each application role. In the example below, Enterprise Applications would be offered Gold and DMZ hosting services, whereas Vendor Applications would be offered Silver, Bronze, and Off-Site hosting services.


After you have the cost model defined, then you simply use some of the attributes in vCenter. The first attributes are going to be the overall CPU count and Memory Size allocated to the virtual machine.


Next, we are going to use some vCenter custom attributes. The custom attributes I created in vCenter include the Build Date, Cost Center, Department Vice President (nothing brings accountability better than attaching the cost to a VP), VM owner, Service Class, and VM Recertification. With these custom attributes, we have a wealth of information for tracking costs, understanding month-to-month build rates, anticipating growth trends, and ensuring that virtual machines are recertified on a yearly basis for lifecycle management.

By creating monthly measurements, I knew that my company was averaging between 30 to 35 new virtual machines a month. This information was critically important to ensure that I was ahead of the curve for resource capacity to meet our business needs and budget forecasting.


To create a Custom Attribute, simply go to Administration in the vCenter console and click on Custom Attributes...



Then click the Add button and create your custom attribute to associate your information with the resources in vCenter.


After you have the information populated in vCenter, you can then export the information from vCenter to a Microsoft Excel Spreadsheet and use formulas to calculate the annual cost. This can be sorted by any of the custom attributes.


Of course, this is a very simple solution, it doesn't include enterprise level core components for self-service management, IT service catalog, capacity management, chargeback system, and lifecycle management. But, all those components come at a cost, and there are a lot of small to medium size businesses that can use the extended capabilities available in vCenter to drive business value.

If you are interested in more information in my cost models, feel free to send me an e-mail.
News: Top vBlog 2016 Trending: DRS Advanced Settings