While technology investments are being made with ever increasing attention to efficiency and return-on-investment, lines of business are demanding more of their IT as a basis for differentiation in fiercely competitive environments. The key to addressing these opposing demands for efficiency and capability is a well-designed infrastructure. Foundational technologies that underpin an enterprise’s applications, from servers and storage up through service buses and portal frameworks, are make-or-break for utilization rates and administration costs on the efficiency side and performance, reliability, and agility on the capability side.

The Java Enterprise Edition (EE) application server, that stalwart runtime “container” that hosts application logic and provides APIs and services for different programming frameworks, integration standards, and capacity scaling, plays a privileged role in the foundation stack. It operates at a low enough level to have visibility into and understanding of fine-grained concerns such memory usage and task parallelization. Simultaneously, the application server operates at a high enough semantic level that it can take into account application-level phenomena such as user response time and state data in making optimization decisions. From this advantageous position, the application server uniquely affords the opportunity to reconcile demands for efficiency with demands to rapidly build, deploy, and adapt competitive capabilities.

Server virtualization using hypervisor-based technologies has significantly improved hardware utilization over the last few years. By allowing workloads previously running on separate dedicated stacks to be consolidated onto shared hardware, hypervisors can increase efficiency while preserving desirable isolation. However, workloads change over time, and how applications are allocated across shared infrastructure must be adjusted to maintain service levels. Server virtualization provides some of this, but with the hypervisor’s limited understanding of the workloads it manipulates, there is only so much optimization it can deliver.

Application servers provide an alternative mechanism that can be exploited for efficiency: clustering. An application server cluster consists of multiple instances of the application server, each with its own copy of the application logic. These instances can contain replicated copies of application state used to support high availability with automatic failover. They can also be used in grid-style parallelization: Application work gets divvied up and load-balanced across the nodes in the cluster. In an environment with multiple applications of varying workloads where workloads spike at different times, clusters can be expanded and contracted across a shared pool of servers, achieving higher utilization and efficiency than if each cluster were statically sized to its respective peak load. The improved utilization achieved with dynamic application server clustering is completely independent of server virtualization, but it arrives at a similar result. Each approach has strengths and weaknesses favoring different environments.

Previous Page Next Page