In my last article, I wrote about how integration is impacting data centers. From a technology perspective, the integration of server, network and storage technologies into a virtual environment is changing how we deliver services to our customers. When we get it right, this new infrastructure gives us massive performance improvements with faster response to the demand for new services.
Then we hit the wall. How do you start to bill for these things? Most data center managers are chartered with creating a rate card in a catalog of services to assist users who are planning their budgets and increasing their knowledge of what future expansion will cost them. Commingling technologies has made it difficult to give the true cost of a fully functional server with all its support equipment. For now, it is probably acceptable to rely on the traditional assessment methods. They can be slightly modified to reflect the virtual impact.
However, going forward, we are going to need a new approach. I propose that it needs to be based on service elements. We are not providing hardware and software. We are providing services. Those services need to be defined financially and then allocated based on a common metric.
I know, that sounds a bit confusing. The following is a general example of a model for service elements and their pricing. (This doesn’t go through all the details, but it gives you insight into the possibilities. ) I will warn you up-front: Creating a model like this one is not a trivial task. It took my team and I almost 18 months to build one. The end result was a huge benefit in planning clarity of billing to the end users.
What are the Service Elements?
First, pick a structure and decide how deep you want to go. Second, identify the billing element. Finally, apply the costs across the model.
A server includes a number of elements. Some of those are:
- The CPU
- The physical device
- The NIC
- The memory
- The chassis or frame
- The rack space
- Power, HVAC and other facility related costs
- OS and software licenses
The beauty of this type of approach is that you choose a level of specificity that will allow you to price in a hybrid environment where you have both virtual and stand alone equipment.
Note, that I stopped at the NIC that is associated with the CPU as far as network goes. This same structural approach should be done for network, storage, security and any other discipline operating within you data center. You need to establish very clear rules among the disciplines to make sure that elements are not duplicated between structures. At the end of the day, all of the service elements will create a total view of your data center delivery costs.
The Billing Element
You have to bill based on something. Will it be the NIC, the CPU or some other element that all the costs will be divided against? Let me make up some numbers so that it is clear I am not using any particular vendor equipment. Assume you have a virtual environment with a physical blade that supports 4 CPUs. That blade sits in a chassis with 5 other blades. You can fit three chassis in one cabinet.
Your CPU cost is made up of the following service elements:
1 – CPU
1/4 - Blade
1/6 – chassis
1/3 – cabinet and associated power and HVAC
See how it works? Now you can layer software costs and maintenance/support costs. You can even add in the average cost of Service Desk support per server CPU if you have those details. The more you can reflect in this space, the more complete your recovery will be.
The tricky part will be when we start providing dynamic CPU service and those measures change throughout the billing period. The beauty of the model is that you can allocate the CPU as a percentage of time over the billing period. The dynamic use then becomes a factor to multiply the base CPU cost by.
This is pretty straightforward math. The physical hardware costs will follow the service element association I outlined above. The blade cost per CPU is 25% (1/4) of the fully loaded blade and so on.
When it comes to the costs like software or maintenance, divide the cost down to your billing element. There are various types of license agreements. Even if you don’t use all the allocated licenses, build the model based on the maximum. Remember, this is not a usage bill, although usage is part of it. You are recovering for your entire infrastructure.
I suggest you start with a subset of your data center infrastructure. Choose technology that is going forward, not legacy. Once you have built the model around that subset, you can expand it to the rest of the infrastructure.
I’d be happy to hear from you on this approach. It has been successful for me in the past, but it can be hard to get your mind around.
To get more useful data center management strategies and insight from Nemertes Research download the Q2 Data Center Knowledge Guide to Enterprise Data Center Strategies – Volume 2.