The following is an excerpt from Visible Ops Private Cloud: From Virtualization to Private Cloud in Four Practical Steps by Andi Mann, Kurt Milne and Jeanne Morain.
To master private
Start by highlighting the cost, service quality, agility, and landscape measures for each environment. Create an elevator pitch that goes something like this:
"We can offer you several options. We can manage your application on a dedicated physical server for $X per year with Y service levels, and it takes Z days to set up.
Or we can put that application on a virtual server. We'll only charge you $X per year based on how many resources are allocated, with Y service level. And it only takes Z hours to set up.
Or we can tailor business-optimizes services in the private cloud. That's only $X per year based on actual resource usage with Y service levels. The really great thing is you can self-service provision it in about 15 minutes. And we can automatically adjust and charge for "just enough" resources as usage levels change.
At comparable levels of resources allocated, it costs more per year in a virtual static environment than in private cloud, but you can fix the resource levels and pay for what is allocated. It costs less per year in a private cloud environment for equivalent usage, but we rightsize resources dynamically to maintain service-level agreements."
This kind of pitch helps position IT as a department that is proactively deploying capabilities designed to quickly respond to changing business needs. As you continue your discussion with business and application owners, emphasize the fact that IT:
- Offers self-service provisioning to match or exceed services offered by third-party providers
- Provides services that were developed with input from business users, and that are tailored specifically to address business needs
- Includes stronger security and compliance controls than are available from public cloud providers
- Scales resources automatically to minimize disruption from changing work levels
- Offers lower annual cost per server by increasing workload density to optimize utilization without sacrificing performance and agility
The value of private cloud
Don't forget, however, that there are costs associated with the tools and automation routines that enable private cloud services. How can you offer agile services at a lower per-unit cost?
The cost efficiency of private cloud computing depends on economies of scale, higher workload density, and dynamic management of resources that ensure service levels. Optimal costs are achieved when all the resources in a pool can be allocated (or even overallocated) to specific workloads. That requires high efficiency in allocating the resources and effective strategies for responding when workloads consume all resources assigned to them.
Conceptually, utilization and agility have an inverse relationship, that is, higher utilization reduces agility. (Think rush-hour grid lock on the highway as compared to 3 AM open-road driving). In a static virtual environment, it's generally true that higher utilization translates into lower agility. In a private cloud environment, however, dynamic resource management strategies that automate response to changing usage levels allow both high utilization and high agility. Automatic response alleviates the queuing and gridlock of oversubscribed resources.
There are fixed and ongoing costs associated with deploying and maintaining the automation and tools that manage dynamic resources in a high-density computing environment. The benefits, which include lower cap-ex and op-ex, must exceed the costs of monitoring, automation, and tooling to justify moving to the private cloud. Otherwise it would make more sense to simply stock a bank of unused virtual servers so you can quickly respond to business needs.
You can use an economic model to show the cost advantage of moving to a private cloud. For example, you can create a model that highlights the footprint measures for the three computing environments -- physical, virtual and cloud -- for both your current data center configuration and the configuration of the target private cloud:
Estimates for physical resource utilization in optimized virtual environment suggest that server, memory and NIC bandwidth can be 70% or higher. However, there are considerable variations based on type of workload, and whether they are deployed in development, QA or production environments. Other estimates suggest that 40% to 60% utilization is typical for highly virtualized environments. Some the IT executives we studied suggested that utilization of 50% or more is required to reduce the cost of cloud services below comparable virtual environment costs.
So how much will private cloud cost?
For cost estimates for each computing environment, many IT organizations charge back virtual servers at half the annual cost of physical. Analysis of cost comparison of private cloud and public cloud suggest that private cloud is 40% less expensive than public cloud for enterprises with significant IT resources already in place.
You can then highlight the current percentages of total IT budget spent on each environment in the current configuration and show how that overall mix changes with private cloud. Another variation shows, for each environment, the percentage of workloads in development, test, and production, and highlights the cost differentials for each environment.
Overall, emphasize the "faster, better, cheaper" value proposition. Highlight custom designed services and greater business alignment in addition to cost comparison metrics.
As you create a cloud model to fit your organization, keep the following in mind:
- Agility is one of the key benefits of a private cloud. So make sure you don't sacrifice agility as you work to drive up utilization.
- Private cloud economics require high utilization to work in your favor. Make sure your service-design process is effective (build the right thing) and monitor customer satisfaction to identify drifting requirements.
- Application targeting must be accurate. Otherwise, you fill your private cloud resources with workloads that are a better fit for a virtual static or physical environment.
For a continued sneak peek at the book, visit the publisher's website.
This was first published in April 2011