Analyzing Potentially Movable Workloads for the Cloud
Few organizations can or should migrate every application to the cloud. You’ll need to identify which ones are the likely candidates, based on such factors as architecture, regulatory constraints, security concerns, and nature of the workload.
Let’s refer to these candidate systems and environments as potentially movable workloads (PMWs).
Once you’ve identified the PMWs you’ll target, your goal is to understand and model all current costs involved in acquiring and operating them. Here are some key steps to constructing a total cost of ownership financial model:
- Reviewing current systems expenditures with key stakeholders and process owners
- Determining appropriate assumptions to be used when modeling indirect costs, such as what portion of network infrastructure costs pertain to the PMWs you’ve identified
- Assessing the likely cost of running key systems in the cloud instead of in your own data center
- Working with stakeholders to confirm assumptions and scope of the modeling effort.
Focusing on just the potentially movable workloads means that certain major cost categories within IT as a whole will have no bearing on the procurement, support, or operation of the potentially movable workloads, and should thus be excluded. For example, if our view of PMWs relates just to server environments, we need to take care to exclude desktop productivity software and support costs from our model.
In addition to costs that need to be excluded entirely, there are data center costs where just a portion pertains to the PMWs you’ve identified (and the portion will be determined through some sort of informed estimate/assumption). These kinds of costs include:
- Network infrastructure
- Data center facilities (depreciation, power, core infrastructure maintenance)
- Building insurance and taxes, building depreciation
- Training & travel.
You’ll probably want to have different areas of your model pertain to different system categories (e.g., dev-related PMWs may require different assumptions from test-related PMWs). For each category, the idea is to determine and/or estimate such factors as total number of servers, labor component for upkeep, depreciation and planned lifecycle maintenance costs, allocated costs for storage, power, cooling, and network components, etc., all rolled up into a “monthly cost per server” estimated total.
What’s the end result of all this hard work? You now have insight into where your current costs come into play, and how those costs would likely change by migration of selected workloads to the cloud. You can now move on to the equally complex issues (and costs) of retooling these applications for cloud operability, and keeping track of the constantly changing pricing and structure of the various cloud offerings available to you.
Peter Kretzman is a consulting CIO and author of the blog, CTO/CIO Perspectives: Intensely practical tips on information technology management.
For more thought provoking cloud management insights visit vmware-erdos.com.