NLANR Network Performance Advisor Demonstrated at Fall Internet2 Meeting

What if your very own network engineer was available 24 hours a day, seven days a week, to analyze and troubleshoot the network performance of your applications at a moment's notice?

That would require impossible amounts of patience and dedication, not to mention infinite quantities of caffeine and pizza. However, Jim Ferguson and the NLANR Distributed Application Support Team are developing what could be the next best thing: NLANR's Network Performance Advisor. The Advisor integrates several existing diagnostic tools into an open-source application that measures, displays, and analyzes network metrics while allowing users to troubleshoot their own networking problems.

"We're trying to make the Advisor a 'junior-level network engineer' that would point out problems that can be analyzed and provide possible fixes in some cases," explains Ferguson.

In mid-October, the Network Performance Advisor was one of a handful of applications demonstrated at the Fall Internet2 Meeting. For the demonstration, NLANR/DAST collaborated with the Internet2 End-to-End Performance Improvement Performance Environment System (piPEs). Both the Advisor and piPES collect measurement data between two networked computers using existing tools, such as ping, traceroute, Iperf, and Web100.

However, while piPES is more of a "middle-of-the-network" diagnostic tool, the Advisor is geared toward end-users, analyzing the network for them and providing solutions in the form of plain-text advice. At SC2003, the Advisor will be analyzing network performance metrics gathered by piPES on nodes within the Abilene Network for their poster display and in the Alliance booth demo.

How does the Advisor work? Suppose you want to transfer data or display visualizations on a remote system across the country using a high-speed network, but for some reason you're not getting the performance you expected. Ferguson says that a user could run the Advisor at either end to collect all the information it could find about the network. "It would then come back and advise the user that there's a problem--a duplex mismatch, for example, and here's something that might solve it," Ferguson explains. "Or it could diagnose a problem that you might not be able to solve, but that your network engineer could." It could also provide a set of diagnostic tests or let the user know that that the network is actually running at optimum efficiency, letting the user know to focus on the application instead.

According to Ferguson, the Advisor's own data collector component will be released at SC2003 in mid-November, with a fuller, more-enhanced version that works with piPES to be demonstrated at the Internet2 Joint Techs workshop in Hawaii in January 2004. The target full release date will be sometime in Spring 2004.

	The Performance Data Collector gathers network performance data via various NLANR measurement and diagnostic tools, which is then analyzed. Data may be accessed via three different user interfaces: the Analysis Interface allows users to interact with the Analysis Engine, which provides plain-text advice and suggestions for troubleshooting; the Expert Interface gives users access to the metrics tree/table; while the Map Interface provides a graphical display of the network. [click thumbnail for larger image]
	The Network Performance Advisor GUI allows users to easily access specific metrics list via tree (left) and table (right) displays and measure them and save or clear settings (buttons at lower right). Connection and security settings can also be accessed via user interface (top). [click thumbnail for larger image]