In the wake of IBM's zEnterprise announcement I've seen and heard many different reactions. On one end of the spectrum is "ho hum, another attempt to make the mainframe relevant." On the other end is "why would anyone choose zEnterprise as a hardware platform when the whole point of cloud applications is that users don't know what's under there."

Those extremes miss the point, as extremes often do.  From my perspective, users are not the direct consumers of these large server platforms. Users consume application services -- users shouldn't care whether the hardware platform is commodity, proprietary or hamster wheels.  The people delivering and operating application services (be they ‘as-a-service' providers, enterprise IT organizations, or outsourcing companies) are the real consumers of these large server platforms.  The people delivering and operating application services will care about hardware platforms if, and only if, the platform provides them competitive advantages.

Competitive advantage arises, in part, because not all application services are the same. This sounds like a fairly obvious statement. However, I still get the sense that many people have not internalized what that obvious statement means for their business and IT organizations.

It means that we will always have computing heterogeneity because different organizations, and different parts of very large organizations, have different goals, types of work, service quality expectations, governance requirements, etc.  For example, a customer billing department does not have the same goals, service expectations and governance requirements as product marketing or even customer support. An online gaming business does not have the same business goals, service expectations and governance requirements as a manufacturing plant or a government agency or a currency exchange.   Matching those differences in an optimal manner often requires application services to have different software architectures, resource configurations, infrastructure resources, tuning expertise, problem diagnosis workflows, etc.  Competitive advantage would accrues to organizations that can rapidly and optimally match data, software, and hardware  capabilities and operational automation to deliver differentiated application services.

From my perspective, IBM's strategy with zEnterprise is about improving the competitive advantages for particular classes of customer with specific types of application services.  For example, organizations that deliver application services which have mainframe data serving as a key component of a tiered or composite software architecture.  Organizations that deliver many application services which have vastly different ways to use a large, common set of structured data.  Organizations developing new application services to connect enormous stores of highly-regulated structured data with event-driven semi-structure or unstructured data.  These organizations (be they ‘as-a-service' providers, enterprise IT organizations, or outsourcing companies) would choose zEnterprise as a platform if the competitive advantage is clear to them.

What IBM is doing for providers of those types of application services, is no different from what HP did with its Extreme Scale Out Data Center Solution for the Google-like service providers of the world. Google's staff didn't build their search engine on top of a relational database, nor does their ‘commodity' hardware look like anything you can get off-the-shelf. To service this type of ‘commodity server' market, HP turned mass production on its head. Instead of mass producing a generic server blades for all service providers, HP started with understanding the needs of a specific class of service provider.  Then HP re-engineered its manufacturing processes and its parts supplier relationships to enable just-in-time delivery of hundreds of server nodes to meet the needs of that class of service provider.  From my perspective, HP is providing just-in-time delivery of hardware designed to match specific types of application services with unique business goals, service expectations and governance requirements.

Application service needs are in the driver's seat for hardware delivery -- and again I stress that not all application services are the same.

Therefore, the point of IBM's announcement is not to make mainframes relevant to ALL types of applications. From my perspective, the big deal about zEnterprise is that IBM is providing a package of heterogeneous hardware and firmware designed and optimized to match specific types of application services with unique business goals, service expectations and governance requirements.

The point is to have the hardware platform improve the competitive advantage obtained from specific types of applications.  Not every IT organization or service provider will have those application needs. Nor should they, if they mean to be competitive!

We are living in a era where business models and competitive advantage are predicated on optimal delivery and operation (dev/ops) of differentiated applications, services and processes.  Part of delivering and operating optimal application services depends on how unique software architectures leverage unique capabilities provided by hardware.  From my perspective, the difference between the HP and IBM solutions mentioned above is how each vendor has decided to deliver hardware ‘as-a-competitive-advantage' for different types of application services.  The choice between the options should become clearer because the starting point for selection is the needs of your specific application services.

• Ptak/Noel Research: 2009 Mainframe Linux Workloads Survey