Fluke Networks SuperAgent 5.0: Testing for the Real World

Application Performance Monitor helps verify vendors' network-performance claims.

May 9, 2004

3 Min Read
NetworkComputing logo in a gray background | NetworkComputing

Individual resources can be grouped in several ways. You can use almost any criteria to classify servers logically or functionally. For example, WAN links may be grouped by region or connection type. This capability will be welcome in a distributed environment with multiple WAN links.

Good

Detailed overview of network response made easy for large environmentsAutomated incident generation and reporting starts the troubleshooting process

Bad

Yet another Windows server to patch and secureManagement interface uses standard HTTP

SuperAgent Application Performance Analyzer, starts at $34,500. Fluke Networks, (800) 283-5853, (425) 446-4519. www.flukenetworks.com

Metrics Made Real

To see how well SuperAgent 5.0 would handle a network of several thousand nodes running a cornucopia of applications, I put it to the test at NETWORK COMPUTING's Syracuse University Real-World Labs. After strategically placing SuperAgent 5.0 to monitor the appropriate spanned router port, I set the product to measure the performance of several applications, including secure Web traffic. I configured reports to be e-mailed as events warranted, but you can also have them trended for historical analysis and reported via broader management platforms, such as OpenView. The product consistently discovered underperforming servers, as well as network segments that deviated too far from baseline measurements.

To test SuperAgent's service-level agreement monitoring function, I first defined performance objectives and translated them into SLAs. For testing purposes, I indicated that 90 percent of my HTTPS traffic to a given server group needed to have a transaction time of no more than 200 ms, and that 98 percent could be no slower than 300 ms.

When performance fell beyond this agreed-upon service level or strayed too far from baseline measurement where no SLA existed, SuperAgent 5.0 launched its own incident investigation. It then dispatched an e-mail to let me know, among other things, which server was deviating from baseline and which application's response times were lagging. Pretty cool stuff-especially when I start thinking about SLAs for everything from our PeopleSoft applications to our Network News Transfer Protocol services.

Overall Goal: Improvement

SuperAgent 5.0 takes much of the guesswork out of high-level network-performance monitoring. It translates gathered data into concise reports and graphs that reflect such critical information as application and server response times, data transfer time, retransmit delays, and network round-trip times. Time samples are user-selectable.

The product offers a vast array of information without overwhelming you. This is no small feat when a large enterprise is being summarized in detail.

My one reservation is that the device is built on Microsoft Windows 2000 Server. With the proliferation of high-yield worms and viruses that target Windows boxes, I'd hate to think this critical management platform might be vulnerable, as it could otherwise prove most beneficial during an attack that's waylaying network performance. Because the management interface uses standard HTTP, the box is particularly susceptible to intruders.

Lee Badman is a network engineer at Syracuse University. Write to him at [email protected].





Among the SLA-oriented details that SuperAgent 5.0 can present are:

  • Packet-loss percentage

  • Connection time

  • Network round-trip time

  • Effective round-trip time (network round-trip time + retransmission delay)

  • Server response time

  • Transaction time

  • Open sessions

  • Data transfer

  • Complete sessions

  • Refused sessions

  • Timed-out sessions

SUBSCRIBE TO OUR NEWSLETTER
Stay informed! Sign up to get expert advice and insight delivered direct to your inbox

You May Also Like


More Insights