* A utility model requires distinct "tiers" of service
On the road to a flexible data center, there are four major steps: consolidation, standardization, virtualization and utility (see “Four steps to utility”). Once a company has implemented virtualization, the final step is to create a utility model. Although a utility model is often seen as equivalent to an outsourced model of “metered computing,” it can be a very useful model even without the outsourcing. Data center managers can use a utility model internally to deliver metered or tiered IT services to the business units. In this context, a utility model provides cost transparency and efficient allocation of IT resources.
Storage is once again one of the first IT services to be converted to a utility model in most organizations. For starters, it is easy to define the “unit” of service delivery: a chunk of storage measured in the megabytes or gigabytes. Since a gigabyte is a gigabyte regardless of how it is used (unlike say computing where Linux CPU time could not be used to run Microsoft applications), the concept of storage as a service can be easily explained (see “Storage leads the way to next-gen data center”).
To implement a utility model, one of the first tasks is to provide distinct “tiers” of service with associated service-level guarantees. At its simplest, this could be a classification such as silver, gold, platinum, defining three tiers with increasing “quality.” How we define the quality in terms of service guarantees depends on the type of resource. For example, in storage, “platinum” may imply synchronous replication vs. asynchronous replication for “gold.” Data center managers need to work closely with their business users to determine which characteristics of IT are important: reliability, availability, performance, latency, jitter, security, etc. Any and all of these characteristics can then be used to define tiered services.
Each tier of service will also have an associated IT cost, which then translates into a pricing level for the end-user of the service. Data center managers can use a modeling technique such as a total cost of ownership (TCO) analysis to determine the full cost of delivering resources as a service. For example, in the case of storage, you can price each gigabyte to reflect the cost of acquisition, operations, backups, depreciation, etc.
However, you must keep in mind that even though you may have different prices for different tiers of service, that does not mean that you have to implement the tiers differently. It may be more cost effective to have a single technology standard, for say storage, regardless of whether you will deliver it as a “silver” or “platinum” level service. The differentiation between tiers could be entirely attributed to operational and other such costs, even though the hardware and architectures are the same. Simply put, in many cases it is cheaper to provision everything on “platinum” hardware, because implementing three different architectures is too complex and will become an operational nightmare. Think of the tiers as virtual contructs that do not necessarily represent different level of technology but instead represent different levels of TCO.
In next week’s article, Nemertes will take another look at IT as a utility and look at transitioning from internal utility to outsourced utility models. Stay tuned.
*** For more about the future of the data center, don’t miss Network World’s six-part series on The New Data Center. On Aug. 22, Network World will publish part 5 in the series, looking at advanced IP developments. On Oct. 4, we publish the part 6 in the series, focusing on outsourcing, new data center-style,Click here to find archives of the past issues in the series.
including an article by Andreas Antonopoulos that provides a layer-by-layer guide to the hottest new data center tools, aimed at application virtualization to intelligent networking.
