Just What Is “On Demand,” Anyway?

Just when it looked like it was safe to talk about IT again, several large vendors have put their weight behind one of the buzziest buzzwords to come along since “dot.com” was on everybody’s lips — and that buzzword is On Demand. On Demand (OD) is particularly buzzy because it is a broad umbrella term that covers several additional buzzwords — utility computing, grid computing, autonomic computing, adaptive management, and others. Vendors using several terms in this family of buzz include IBM, HP, Computer Associates, Microsoft, and Sun Microsystems, to name some of the more prominent examples. Fortunately, ZapThink loves debuzzifying terms like these. Naturally, we’ll provide the Service-Oriented perspective on these terms, but first, let’s introduce some clarity.

The story starts with one core business problem that faces any organization with a large number of computer systems: the supply of computing power is inelastic, while the demand for that power is elastic. In other words, companies must either run a large number of systems to handle peaks in demand, leaving those systems mostly idle the rest of the time — which is a costly use of scant IT budgets, or run a smaller number of systems that meets the companies’ needs most of the time, but fails occasionally under high demand — which is even worse. Neither option is very appealing.

Pieces of On Demand
While OD is a somewhat nebulous concept that centers on agile businesses that take advantage of flexible IT infrastructures, the component concepts that fall under the OD umbrella have more specific meanings. Utility computing is a combination of two approaches for solving these problems. The first approach is a managed operations delivery model, where companies can call upon a third party to host and manage their IT infrastructure, thus providing access to extra resources to handle those peak loads. The second approach is a pay as you go financial model for obtaining IT capabilities, where companies can then pay for the resources they use following a metered services model that changes a fixed IT cost into a variable cost, which is easier for many companies to swallow.

Another, related definition of utility computing is the shared pool approach to IT resources. Companies can centralize their IT infrastructure into a shared pool of resources, providing “use it when you need it” functionality in-house to various users across the enterprise, giving them the benefits of utility computing without the need for a third party provider.

Grid computing is similar to utility computing in some respects, but takes a different approach to solve a different set of problems. Grid computing is a form of virtualization — taking a large number of systems and combining them into one massive, “virtual” computer, or grid, that can handle computation-intensive tasks. Such grids can include widely distributed systems or systems within particular data centers. Grids are in use today in fields as diverse as genomics and massively multi-player gaming. Grids can also provide a way for IT organizations to put idle system resources to use, solving one of the problems that utility computing addresses.

The virtualization of computing resources is itself a concept under the OD banner. Virtualization of storage and application functionality as well as the virtualization of systems enable the pooling of IT resources that form the core of both grid and shared pool utility computing. Another enabling set of technologies goes under the names of self-management, adaptive management, or autonomic computing. These concepts are some combination of technologies that help systems recover from problems automatically — essentially, a “self-healing” capability — as well as system management capabilities that are inherently flexible, in order to support the dynamic needs of utility and grid computing.

What Vendors Are Doing
Naturally, every vendor who has tossed its hat into the OD ring has a different twist on the problem. Here’s a quick overview of what the better known players are doing:

  • Sun Microsystems’ N1 platform virtualizes systems and storage resources, providing the ability to provision computer, storage, and network resources on the fly.
  • HP’s Utility Data Center and adaptive management infrastructure built on OpenView aims to interlink systems, networking, and storage to form a grid infrastructure.
  • CA has announced a set of offerings under their Unicenter brand that enable resources to be managed beyond the enterprise in a grid of computing power.
  • Microsoft has launched its Dynamic Systems Initiative, which unifies hardware, software and service vendors around a software architecture that enables customers to harness the power of industry-standard hardware.
  • IBM has focused their entire company on the “eBusiness On Demand” vision, which is a conceptual framework for determining business needs and transforming companies to take advantage of utility computing, grid computing, and other technologies that can increase the flexibility and responsiveness of businesses.

Sun is taking a virtualization approach to OD, while HP, CA, and Microsoft are offering adaptive management capabilities to enable companies to build and operate OD infrastructures. IBM has the most comprehensive vision for OD, including business consulting and transformation along with the concepts of virtualization and management as well as grid and utility computing under the OD banner. If you’re confused about how to compare these vendors’ offerings and approaches, you’re not alone. Unfortunately, there is no one common definition of OD computing that might provide some clarity to the marketplace. Instead, we have this mishmash of related concepts — utility computing, metered/pay as you go, shared pool, grid computing, adaptive management, virtualization, dynamic provisioning, and more.

The Unifying Vision
A large part of the confusion over OD is due to the commingling of business and technology concepts. On the one hand, OD has to do with how companies pay for and use IT, and on the other hand, OD talks about how to manage and coordinate systems and other resources. What we need is a way of thinking about IT resources that separate these two realms so that business needs can drive technology decisions in a flexible, agile manner. ZapThink believes that Service Orientation provides the missing conceptual link.

Service Orientation is a way of looking at IT that abstracts the functionality of the technology into discrete, business-oriented chunks, known as software services. From the Service-oriented business perspective, IT doesn’t consist of servers and networks and applications. Instead, IT provides services — functionality that can be called upon and incorporated into business processes as needed.

Taking a Service-oriented view of OD helps to clear up much of the confusion. The role of IT is to provide and support the services that the business requires, but just how it provides those services depends on the characteristics of the particular implementation. If it makes sense to pool servers or outsource functionality, so be it — as long as the business has access to the services it needs. The services therefore become a kind of conduit between line of business and IT — focusing and providing context for business requirements for IT personnel, while at the same time hiding the complexity of the implementation details under a useful layer of abstraction for line of business users.

The vendors mentioned above should take note: your various OD messages are confusing your customers. You need some kind of organizing message that clarifies the business and technology benefits of your initiatives for the appropriate audiences. Service Orientation is that organizing message.