Next Generation Network Series: Utility Computing and Business Value

The first step toward increased flexibility is utilisation-based pricing. This model still involves a fixed number of dedicated servers, but instead of a flat rate, customers pay a variable rate based on how much they actually use these servers. Run them lightly, and pay less; then pay incrementally for increasing loads. An even more flexible approach is true utility pricing, where resources are dynamically allocated according to demand, and customers are only charged for the number of resources allocated for a given amount of time. IT managers tend to build for peak demand, on the principle that underestimating requirements can lead to performance problems, which can then lead to unhappy customers.

They prefer separate dedicated resources in their environment to handle eCommerce, payroll, year-end accounting, email, ERP and so forth. But the truth is that these servers rarely run at full capacity, even in extreme conditions. Most studies show that typical average server utilization is around 10%. That means most companies have far more dedicated resources than they need, because they’ve been acquired and deployed to support a worst-case scenario. In fact, a blended resource environment makes the most sense. Dedicated servers on flat-rate leases are still the most cost-effective way to handle daily baseline demand. Usage-sensitive servers can be added to the mix to support typical variations in demand due to normal statistical fluctuations. Pay-per-use servers might seem to be the most expensive option when considered by per-minute price, but if they’re part of the portfolio, they can be used to handle unexpected spikes or other unusual situations. They actually save money, since a premium is paid for the short time those resources are used, but nothing is paid when they’re not used!