Virtualization: must include application services as well as resources.

In an economic climate where every IT investment is closely scrutinized and companies face increased pressure to do more with less, IT expenditures must be aligned with overall business objectives and tied to quantifiable results. Today, global corporations are staggering under the burden of a complex and costly IT infrastructure that is characterized by a horde of idle servers and incompatible applications--and no elegant way to get the "old" to talk to the "new."

In response to this never-ending strife between business users and IT, the approach to technology architecture and infrastructure is experiencing a fundamental shift--the likes of which haven't been seen since the advent of client/server capabilities. Organizations are working to create a virtual infrastructure that automatically shares and manages computing resources across all applications within the enterprise.

The principles of virtualization (the notion of liberating functionality and management functions from dedicated physical devices) have been around for decades. One example is distributed computing, more recently known as grid computing. Grids apply the concept of resource virtualization by aggregating numerous compute resources (from mainframes to desktops) and abstracting them into a single, unified resource that provides, in effect, a single virtual computer that can be accessed as needed or on demand.

Whether internally or externally provisioned, utility or on-demand computing allows IT organizations to meet service-level agreements in an extremely efficient and economic manner, by leveraging heterogeneous hardware platforms. Typically, applications housing critical business logic are built and deployed in some form of dedicated silo; this creates a costly dependence on resources. Often, business demand simply cannot be met, or it cannot be met at an economic level. Virtualization eliminates 'islands of processing' and optimizes the use of resources through a common, shared platform available to applications by location, business unit, operating company and even the whole enterprise. In this model, the business can exploit resources effectively and fundamentally change the cost-growth equation.

[FIGURE 1 OMITTED]

However, that resource-centric view is only a piece of the puzzle. In effect, it addresses only the "supply" side of the equation. But what of business "demand"? Organizations considering a virtualization strategy must also take a "top-down" view of current operations to determine which business-critical applications represent bottlenecks and pain points for the business users.

To create a true virtualized infrastructure, organizations must consider both resource virtualization and application (services) virtualization. While the former represents the more common approach that offers numerous improvements in IT performance, application services virtualization represents a breakthrough that can radically impact both IT and business operations.

Resource Virtualization

Through a software layer in the IT stack, resource virtualization provides an abstraction that allows applications to access physical resources indirectly, without being tied directly to specific physical resources (such as servers or clusters). This virtualization software provides a common interface for applications. There is no limit to the number of resources (limiting scalability) that may be put on a problem. Through load balancing, adaptive scheduling resource monitoring/diagnostics and policy/prioritization management, the software distributes the work across whatever individual resources are available to process the work.

Application Virtualization 

Today, applications housing critical business logic are usually built and deployed in some form of dedicated silo. These silos limit business performance by introducing dependencies in systems capacity, availability, partitioning, run-time load, and more. As a result, the business suffers from latency, failures, or inappropriate use of the underlying resources of the system to process the application workload/servicing accordingly.

In practice, application virtualization is most appropriately called application service virtualization. Application service virtualization allows business functionality hosted as application services to be accessed however required (synchronous, asynchronous, ad-hoc or predictable) without requiring a physical system endpoint to be known or even needed to execute. Application virtualization provides a simple means to accept services to be published and hosted in a framework without requiring cumbersome manipulation or modification of application code or strict limitations on code types.

For example, code component support needs to handle all types, from Java and .NET to C++, legacy code and executables themselves. In addition to a broad range of code support that can be easily implemented in a virtualized environment, there also needs to be a range of simple access methods to request execution of these application services or components. Support should include access methods like rich WSDL-like interfaces, to common access methods like SOAP or scripting to allow any application to leverage a virtualized infrastructure fully.

Application service virtualization helps to guarantee application execution over a distributed pool of compute resources. A virtualized application is not only capable of remotely invoking application requests and returning results, but also ensuring that the application's state and other data are available and consistent on all resource nodes executing the application across a grid.

Once application services are virtualized, organizations can achieve a new model of IT consumption and fulfillment. The ultimate goal is to satisfy the applications' execution needs by balancing processing requirements (e.g., priority, service level, resource requirements, exception handling) with resource coordination (e.g., resource matching, data requirements, service libraries, state management, parallelization of service requires, service affinity of run-time nodes) in a virtualized environment.

This virtualization mechanism allows multiple applications with multiple users the ability to access services in ad-hoc unpredictable patterns along with regularly scheduled daily processing. By disaggregating applications into their constituent components of services, scheduling can become much more dynamic and load balancing more effective. Any types of failure impacts only a smaller portion of a total application or business process (rather than an entire application failure), and through application virtualization's failover capabilities the individual piece of work is automatically redirected to another resource, virtually eliminating any real impact to the business.

In addition to delivering a high-throughput, low-latency messaging infrastructure that underlies communication between application components, an application virtualization (or "service") grid should contain sophisticated adaptive scheduling and secure, robust, fault-tolerant components. Application virtualization infrastructure is an integral part of the enterprise IT infrastructure and must be extensible, standards-oriented and include connectivity to the data center for systems management tools, enterprise administration, monitoring, and logging.

Combining resource and application virtualization is necessary to realize the benefits of a true service-oriented architecture. This virtualized infrastructure allows applications to non-invasively leverage and exploit heterogeneous resources across the enterprise. It also manages the execution of the services as required to meet the service levels and cost structures of the business. Ultimately, by creating a new model which balances demand (business requirements) with supply (available IT resources), virtualization can create an agile IT infrastructure that can optimize business performance and IT costs/investments.

www.datasynapse.com

Tony Bishop is chief strategy officer, DataSynapse, Inc. (New York, NY)