Search this site with

End User Tools and Projects
- NextINet
- Advanced Applications
Database
- DAST Projects/Tools
- Network Performance
and Measurement Tools

End User Support
- Getting Started Guide
- Networking Glossary
- Other Projects/Organizations
- Funding Opportunities

Documents
- Guides/Tutorials
- Papers/Articles
- Presentations
- Reference Books

WebCT Courses
- Tuning Applications

Events
- NLANR/DAST Training
- NLANR Packets Calendar
- Idesk Travel Schedule

News
- Press Releases
- Alliance Data Link
- I2 Newswire Archives

Reports & Statistics
- Monthly Updates and QSRs
- Abilene "Weather Map"
- Web Server Stats

Started in June 1997, Surveyor is a project to address universities' Internet performance issues with relevant data and to develop new approaches towards solving them. Surveyor was developed by Advanced Network & Services. The development team works closely with the IETF IPPM (Internet Protocol Performance Metrics) Working Group with a goal of standardizing a set of metrics for the quality of IP delivery services.

Similar to the NLANR AMP approach, one measurement machine is deployed at each participating site. As of July 1999, there are 55 Surveyor sites, including a few outside the U.S. Some of the sites overlap with NLANR AMP sites. The metrics being measured include one-way delay, packet loss rate, and routing. Unlike AMP, Surveyor requires installing a Global Positioning System (GPS) device on each measurement machine to synchronize the clock to accurately measure the one-way packet delay. The test packets are time-stamped and sent as a continuous Poisson stream to other machines. The one-way delay is computed by subtracting the time-stamp in the packet from the time of arrival at the destination machine. The motivation of one-way delay instead of round-trip-time is to separate the delay of the forward and backward paths, which can have asymmetric latencies.

At the Surveyor web interface, users can select two sites and generate a report for a specific date. Three types of graph plots are available:

• percentile delay that plots 0th, 50th, and 90th percentile points
• delay histogram that plots the percentages of packets versus delay
• percent loss that plots the packet loss rate

Figure 2 below shows an interesting example of one-way delay percentile from NCSA to Harvard. In the figure you can clearly see the delay pattern during that day. The blue points indicate the 0th percentile (minimal) delay, the green ones indicate the 50th percentile, and the red ones indicate the 90th percentile. In an uncongested path, the blue, green, and red points should be close to each other, similar to what is shown close to the 0:00 time. In a congested network, the blue points would likely stay the same because at least one of the packets did not experience much congestion. But the green and red points would "drift" from the blue points considerably. Such drift is a measure of congestion. According to the figure, the path seemed to be getting worse and worse during the first half of that day, and the blue points suddenly changed to a higher value at around 13:30 time. Changes in the value of the blue points typically indicate a change in the path or an extreme case of congestion. In this case, it turned out that the vBNS routing went down during that period, which required packets to take 7 hops from NCSA to Harvard via vBNS and 22 hops via commodity Internet during the vBNS outage. The vBNS routing was back in place later that day, indicated by the changed-back of the blue points. For detail information about interpreting the results, please refer to the online user's guide.

Figure 2. Surveyor one-way delay percentile from NCSA to Harvard

To become a Surveyor site, first read through the procedure, then send an email to surveyor-help@advanced.org.