Maintaining a data center at optimum performance and utilization is not a one-time task. As customer requirements change, the pace of innovation must adjust accordingly.

As a result, data center operators perform periodic evaluations and optimizations to ensure that the infrastructure supports high performance, availability and reliability. The alternative is an underperforming asset with unnecessary costs and the likelihood of business disruption that can result in revenue loss, customer base erosion, and brand damage.

Know the Delta – Evaluate, then Optimize

The goal of data center optimization is to increase efficiency in both performance and utilization. You must first evaluate the current performance of the center to establish a baseline, which allows you to:

• ‑Empirically document current performance

• ‑Measure performance against targets

• ‑Isolate indicators of network deficiencies

• ‑Focus efforts on prime areas for optimization

• ‑Avoid changes unlikely to deliver significant gains

• ‑Reduce the effort and cost of the project

• ‑Measure the degree of improvement

Network Evaluation Best Practices

Core considerations when developing a data-center evaluation strategy include the following:

• ‑Benchmark performance. Have a clear picture of performance and problems before you make changes. After each phase, benchmarks provide an invaluable reference to track relative performance.

• ‑Identify performance metrics. Define the level of performance desired, then establish the performance metrics, such as throughput and latency, and use them as part of the benchmarking process.

• ‑Identify performance problem areas. Know which problem to fix. Is slow transaction response time due to CPU utilization, dropped sessions, overloaded queues, congestion at the top-of-rack switch, or disk access issues? Fix the right problem instead of following a time-consuming trial-and-error approach.

• ‑Measure utilization levels. Understand how much headroom you have before upgrades and optimizations must be made. Plan strategically rather than operate in an inefficient reactive mode.

• ‑Make informed decisions. You may have a range of options under consideration, such as data center consolidation, virtualization or cloud computing. Hard data in the form of network metrics are invaluable in directing you to the best solution.

• ‑Demonstrate the effectiveness of your solution. Measure your project’s success with empirical data before and after test results to determine ROI.

Building the Team

The evaluation team is a critical component of your project. You may have the in-house resources required, including test systems and people with network testing expertise and the time in their schedule to devote to this project. If these resources are not available, it may be worthwhile to engage an experienced third party. Clearly, an important advantage of outside resources is objectivity. It’s not uncommon that when home-grown systems are tested by their developers, operating assumptions may not be discussed or even challenged, and failure points can easily be overlooked.

Look for a testing partner with:

• ‑Experience in the type of testing required

• ‑Testing as a core competency (not as a sideline to implementation)

• ‑Use of test equipment as part of the package (which can often be cost-prohibitive to own)

• ‑A best-practices test methodology

• ‑An established delivery process

• ‑References and success stories

Conclusion

Retooling or optimizing a data center is a key step in maintaining efficiency. Network evaluation and benchmarking are often overlooked during planning and implementation, resulting in time-consuming and costly problems during and after the launch. Data center managers can ensure a smooth launch by scheduling and budgeting for testing early in the process.

Larry Goldstein, Director Enterprise Business - Global Services, Spirent Communications, 310-377-5459; email: larry.goldstein@spirent.com; web: www.spirent.com.