Rich Internet Applications (RIAs) were discussed in my last posting as providing a higher-level abstraction consistent with information virtualization. Virtualization first made a big impact on storage and then evolved to servers followed by applications. Continuing this trend, virtualization is now increasingly focused on the information that provides the rationale for storage, servers and applications.

RIAs are part of this evolution since they enable information access based on controlled vocabulary terms defined for use in taxonomy categories, which removes implementation details and reduces coupling. RIAs are related to Rich Internet Clients (RICs), which are client applications that go beyond traditional browser-based clients by presenting a user interface that’s similar to and largely as capable as what’s offered by native desktop applications. RICs represent the client end of the computing infrastructure while clouds and cloud computing represent the backend designed to store information and host applications. RIAs, RICs and clouds all virtualize information by hiding implementation details, which reduces coupling. SOA also reduces coupling and sets the stage for Information-as-a-Service (IaaS), both of which were discussed in the last posting. SOA requires further explanation while coupling and cloud computing are new topics.

SOA is an architecture style designed to achieve loose coupling between software agents by defining and exposing interfaces as standard service contracts, which are published in public or private shared service registries and invoked using a communications model based on messages defined by schemas. The core concepts for SOA, therefore, are coupling, contracts, registries, messages and schemas. Coupling is useful for explaining virtualization, which is the focus of this posting. Differences between contracts and interfaces, and the properties and purposes of registries, messages and schemas are topics for future postings.

Coupling measures the dependencies between interacting systems, which fall into two categories: real coupling and artificial coupling. Real coupling occurs when one system produces outputs or provides services needed by another system. This is acceptable assuming the scope of and the contractual relationship between the two systems is correctly defined. In contrast, artificial coupling represents unnecessary dependencies that should be minimized or eliminated to provide better service quality. An obvious example involves failure where you want systems to have independent failure modalities. Loose coupling also allows services to be created and evolve independently. Dependencies that should be minimized typically include the identities and names of resources, spatial and temporal relationships between resources, and implementation details such as the internal structure and content of resources and their performance properties such as bandwidth or latency.

Cloud computing is a computing model characterized by real-time provisioning of resources to support on-demand service, ubiquitous network access that’s typically Internet-based, location-independent resource pooling, rapid elasticity and service metrics to support service management and billing. With cloud computing, users need not be concerned with resource provisioning, capacity planning, and many other implementation details. There are three primary cloud computing delivery models that virtualize different elements of the computing infrastructure: Infrastructure-as-a-Service (confusingly, Information-as-a-Service and Infrastructure-as-a-Service have the same IaaS acronym), Platform-as-a-Service (PaaS) and Software-as-a-Service (SaaS).

Virtualization applies to all technology segments, including hardware, software and information, which is why it’s the most significant technology trend impacting IT in many years. In addition to impacting all segments, virtualization can span the transaction path length from end-to-end since clients, networks, servers, storage, applications and information are all primary virtualization targets. Major trends such as SOA and cloud computing are particular forms of virtualization, which illustrates its broad impact and the many ways it can be exploited.

But is there a definition of virtualization that explains what it is regardless of the market segment or implementation details? Three examples illustrate differences and similarities with various forms of virtualization:

• Virtual machines (VMs) – VMs are software implementations of real machines (computers), which execute programs like real machines. By replacing a physical view of a processor with a logical view, programs are insulated from implementation details of the underlying machine such as the instruction set.
• Virtual information – Virtual information is information that’s been virtualized to make it available independent of implementation details, including location, format, access method and storage mechanism.
• Virtual workloads - Grids virtualize workloads by mapping application workloads to available computing resources. Workload virtualization is related to but different from service virtualization because all services represent workloads but not all workloads are services. Workload virtualization requires that the grid solution be integrated with whatever is generating the workloads.

Virtualization replaces physical views of computing resources with logical views, which makes them easier to use through the removal of implementation details, including language, geographic location, format and physical packaging. In addition to storage, servers, applications and information, entire data centers are being virtualized. Cloud computing is enabling the virtualization of complete computing infrastructures that may involve multiple data centers, and we can even expect entire businesses to be virtualized, including the staff and associated processes.