If a new application's performance leaves a lot to be desired, developers often point the finger at network folks, who say the network is just fine. Resolving this type of dispute is the forte of Opnet Technologies' IT Guru - a sophisticated tool that provides analysis and modeling of network performance.

We recently tested Version 9.0 of IT Guru at West Virginia University's Advanced Network Applications Lab. The product performed as expected, and was very user-friendly in its basic features. For companies in which network performance issues can cost thousands of dollars in lost revenue per minute, the package is invaluable.

Strong data analysis

IT Guru's data analysis begins with packet capture. The Application Characterization Environment (ACE) module includes a capture agent that can be installed on various machines throughout the network. This lets you measure performance from various vantage points across the network. The ACE module also can import packet traces from network analyzers, such as Network Associates' Sniffer series.

After visually inspecting traffic, ACE takes the raw data and filters it by various criteria. This is where the performance analysis begins. The ability to view network and application performance side by side in a graphical format is a nice feature, which makes it easy to determine the location of performance issues and delays.

The AppDoctor module offers spreadsheet-like diagnostic and statistic windows that help determine if delays are occurring in the application or network. AppDoctor analyzes different parameters, including average packet size, retransmissions, the number of no acknowledgements, and errors. Any out-of-range values are highlighted in yellow by default. Although you can set the out-of-range thresholds, we found the default settings adequate for most diagnoses. AppDoctor gives basic explanations of the delay types it identifies and offers possible solutions. While far from comprehensive, we found the information useful.

Advanced controls in the Quick Predict module let you introduce variables that can help predict the effect of changes on the network. These variables include changing the bandwidth, latency, packet loss, link optimization, payload/packet size and TCP window size (setting the optimum TCP window size is often overlooked in network performance).