You cannot manage what you do not measure
Like water from the tap or electricity out of the wall socket, computing power seems to flow freely from the cloud. Cloud computing is so (or too?) easy; you can go to one of the cloud service providers, use your credit card to sign up for a virtual server and there you go. A new instance is only a few clicks away. Your only notion of having a server running somewhere comes when the monthly bill arrives.
In earlier days the pricing model for virtual machines in the cloud was fairly simple. The variables in pricing were the amount of CPUs and the storage provided. This has now evolved into a more fine-grain model, and depending on the provider you can compose your own menu: T-shirt-size-based CPU (small, medium, large), storage IOPS (input/output operations per second) and size in GB or TB, network bandwidth and optional clustering capabilities.
What’s next? I can imagine, and would very much welcome, the amount of CPU cycles used during a certain timeframe becoming the base of billing. “Pay per tick”—the most transparent way of doing business. In this way people would also start thinking about resource consumption of an application and developers would be persuaded to make more lean applications (but please, don’t give up security).
We’re spoiled; in earlier days when computing resources were scarce we had to think about every block, byte and clock-cycle, but nowadays resources seem to be unlimited.
In other words, application developers (in house and independent software vendors [ISVs]) can help companies to save costs by developing a small footprint with energy-efficient applications. But this will be a challenge.
Most business application developers I see coming along in the IBM Innovation Center do not design their applications with performance as their primary goal. They want to build a nice-looking, fully functional application. The nonfunctional requirements usually have specifications regarding response times, but almost no hardware requirements are specified. Hardware is not an issue for them—just buy a bigger box. During development all the functional tests are carried out, but good performance tests are scarcely done.
My discussions with these developers are always the same: What design choices did you make, and what are the performance impacts of them? Why do you use framework X in Java to support portability between platforms where in practice the only target is platform Y? I am not against using frameworks in Java; I am only asking that you use them with care and know what the performance impacts are. And usually the developers don’t have a clue about the impact. With that “knowledge” we start testing, compare the results, improve, test and compare again. In this way you get in-depth insight on changes in your application.
In order to be able to bill per tick you’ll need a multiplatform solution that helps you to generate end-to-end IT charging and accounting reports. The IBM Tivoli brand has a product that satisfies the requirements for an integrated and cross-platform solution: Tivoli Usage and Accounting Manager. Different plug-ins enable billing based on using and costs of servers, databases, storage and networks. With it, you can understand your costs, track, allocate and invoice by department, user and many additional criteria.
Having such an infrastructure and processes, all the parties involved can benefit.
How the provider benefits:
- Value proposition: reward savings in resource utilization; distinguish yourself as a “green” company.
- The utilization data can be used as input for capacity planning. Trends in needed capacity can be used as input to justify investments in infrastructure.
- It can lead to lower operational costs. High, normal and reduced rates for designed time slots can smoothen utilization spikes and thus lower investments costs.
How the consumer benefits:
- Pay for what you use; savings are rewarded.
- Cost transparency: more server load of your web shop means more business. On low hours, less will be charged.
In either private or public cloud, the principle stays the same; pay for what you really use, and if you are smart you can save money.
When discussing this topic with one of my peers he asked me, “Do we go back to the mainframe era where we all had this in place?” My response was that, seeing the trend that you have to show the business value of your activities, you need a solution that can help you to make the business value visible. When the billing is fair, understandable and accurate it will work.
Learn more about IBM SmartCloud Cost Management.
Henk Waanders operates at the leading edge of IBM’s hard- and software portfolio. Not only by telling the IBM story to ISV’s and Business Partners, he is also responsible for the hands-on part; installing and configuring the systems needed for demos. Follow Henk on Twitter: @HenkWaanders