OPTIMIZED SYSTEMS

Delivering Business Value With Workload-Optimized Systems

Michael Steinhart | Date: 01-27-12 | Comments

x86 servers are good for general-purpose applications like email and file storage, but higher-powered servers with multicore processors and large memory capacities can turbocharge database and transactional applications while maximizing utilization.

The idea behind workload-optimized systems is simple: Instead of buying and deploying the best and latest technology in order to reap performance benefits, you buy the technology that's best-suited for your workloads, and you make sure each application has the optimal mix of resources for high performance, reliability and efficiency.

You can figure out how to optimize workloads by analyzing the behavior of different applications. Some need large amounts of memory, others read data but don't write any, others require strong security. The mix of servers and processors you choose should correspond to these specific usage patterns, instead of hoping that one size fits all.

A recent advisory from Clabby Analytics indicates that workload-optimized systems deliver more business value than "traditional" computing systems. "The business impact is greatest when the server can execute the work�and deliver the results to the business�in a timeframe that has maximum impact, at no greater cost than is necessary to meet business service-level requirements," said Joe Clabby, president of the firm.

The report goes on to explain that x86 servers are good for general-purpose applications such as email and file storage, while higher-powered servers with multicore processors and large memory capacities can turbocharge database and transactional applications while maximizing utilization.

Now, the goal of workload optimization is to match application characteristics and service-level requirements to the best platform for the job, be it a cluster, an x86 rack or blade server, a grid environment, and so on.

When companies undertake a workload optimization initiative and virtualize an application, they must track all of its instances and software licenses, map its resource utilization and requirements, and designate its security and resiliency needs. When all this is done, they'll have the blueprint for an optimal infrastructure that works efficiently without wasting energy or capacity.

In one example, the report cites an energy company that was able to consolidate 20 physical Sun servers onto two IBM Power Systems servers. The company also reduced its x86 server footprint from 60 hardware boxes to four.

It accomplished this dramatic consolidation by tuning each application and leveraging advanced virtualization features to make sure every processor cycle is utilized to the fullest. With this fine-tuning in place, system performance is actually up by 92 percent, and batch runs that took 24 hours in the past now take only 2.

This consolidation and virtualization project also saves the company $1.1 million each year in operating costs and software licenses, while delivering power and performance exactly where needed.