March Issue of the Taos Newsletter: Utility Computing

Case Study: Evaluating the Effectiveness of Utility Computing to Gain Efficiencies, Increase Services, and Reduce Costs

Taos Professional Services Team

Utility Computing is causing a major shift in how IT organizations are run - moving away from a model where servers are dedicated to particular tasks and towards a model where tasks are run on a flexible pool of resources. Different vendors approach Utility Computing from different directions, and IT organizations have to figure out the best way to apply the competing approaches to their business environment to reduce costs and/or gain functionality.

As a professional IT consulting services company, Taos is called upon to help clients sort out the issues, relying on our experience with the many Utility Computing technologies impacting IT today. These technologies include LSF, virtual servers, clustering, storage virtualization, SAN consolidation, storage chargeback systems, and the general automation and management of very large server farms.

The following case study demonstrates how Taos is helping a customer advance into the area of Utility Computing using blade technology.

The Customer Challenge

This particular Taos customer has many somewhat related customer-facing departments, each with its own internal IT requirements as well as web and other online services it provides to customers. Previously, IT standardized on system platforms and technologies, but let each business unit build its own solution, separate from other business units. The result was that data was duplicated, server capabilities duplicated, and large amounts of dollars were soaked up in redundant efforts.

Utility Computing - A Multi-Level Strategy for the Future?

In order to improve efficiency and reduce costs, the customer decided to evaluate future strategies, including Utility Computing at several levels.

At the very top level, the customer is considering a re-architecture of their business, changing their "islands of information" into "islands of services". In other words, bring the business units together in a framework of shared common internal and customer-facing services so that each department can then tie together with others as needed to support both their business data and their customer needs.

It is expected that this fundamental redesign of how IT provides services to multiple internal (and indirectly, external) customers will enable a long-term shift in how this company can do business - providing higher levels of service at reduced costs. Shared services will not have to be duplicated other than for failure redundancy and total capacity; and shared data can be reused by each department that touches a given customer. The architecture and analysis process to build this sort of high-efficiency, services-oriented network through Utility Computing is underway today.

As a shorter term goal, the customer IT department is also looking to apply a different type of Utility Computing to other levels of their environment - databases and web application servers running Java code. Since these are the most common high performance systems in their production environment, these systems can also benefit from the concept of Utility Computing.

Due to capacity and reliability concerns, the customer has been buying large and expensive Enterprise class servers from major hardware vendors. But today there are smaller, lower cost alternatives on the market that can be connected to provide the same or even better level of reliability and capacity. Many applications can be clustered on these smaller systems, including leading databases such as Oracle, and various web application servers. Oracle's 9i RAC cluster technology can bring a large number of small systems together to replace much larger servers. In addition to reducing costs through using more inexpensive hardware, the customer is also considering a fundamental shift in software - for example, the use of modern Open Software operating systems such as enterprise versions of Linux.

Taos is Helping in Two Ways

Having identified a number of potential hardware and software approaches for future servers, the customer engaged Taos to help in two major areas:

First, Taos consultants helped build the proof-of-concept systems, including operating system and application integration work, and then helped debug the systems and networks. This work helped test and demonstrate a key Utility Computing function - the ability to automatically detect and remove failed systems without interruption to the service.

Second, Taos is helping the customer generate standard performance evaluations across the various platforms under load and failure testing, so that they will have a consistent way to evaluate the performance of the test configurations and judge their cost effectiveness and stability.

To help develop the decision criteria, Taos consultants worked with the various applications and internal customer business units to identify what performance parameters were needed for the business evaluation, and how to gather and present that data during the testing and evaluation. Several separate business lines are involved in the testing, across several applications and across a wide variety of system and software configurations.

An example of one of the most interesting technologies being tested is a hardware virtual server rack system. This server rack, produced by Egenera, creates virtual systems out of a shared pool of processor cards, memory cards, and I/O devices. The cards and devices can be reallocated if needed due to load demands or due to a failure of hardware, appearing to the supported Linux or Windows operating systems as a physical server even though boards and components may be reallocated by the rack's control units.

The Results Look Promising

While this particular case study is still underway, our results so far are very promising. The ability to shift hardware and components where needed, yet still use standard operating systems, is one of the major benefits that is helping to drive Utility Computing. And Taos is helping the customer identify how well it works in terms of performance and reliability compared to today's legacy servers.