This story appeared on Network World Fusion at

'Net Insider:

Jittery about jitter?

By Scott Bradner
Network World, 08/20/01            

We are told in just about every venue that the Internet needs all sorts of quality-of-service mechanisms to make it useful. However, some recent real-world experiments seriously question whether this is true.

The experiments do not indicate that today's Internet is already "five 9s" reliable (the mythical reliability of the phone system) over long periods of time, but did show "four 9s" over the test period and 100% ("infinite 9s"?) reliability for one particular week.

Steve Casner, Cengiz Alaettinoglu and Chia-Chee Kuan of Packet Design conducted the tests, and the results were reported in May at The North American Network Operators' Group (NANOG) meeting. The presentation can be found at here. (For the record, I am on the Packet Design technical advisory board.)

The test was not an easy one. A 1M bit/sec stream of traffic was sent between test hosts installed in an ISP's points of presence in San Francisco and Washington, D.C. The datastream, which went through the ISP's routers and operational backbone links, consisted of random-length packets - between 64 and 1,500 bytes long - at random intervals (with a 6 msec mean interval). The tests ran for 15 periods of five to seven days each. A time stamp was included in each packet, and the latency was measured with 20-microsecond accuracy. The observed jitter during the entire test was less than 1 msec for 99.99% of the packets. The availability was the same - 99.99%.

Even with this level of reliability, a further improvement was made by changing things so that the routers' address resolution protocol tables did not timeout. With this change, and an absence of "fibertropic backhoes," 69 million packets were sent over a week with zero packets lost and 100% of the jitter less than 700 microseconds.

There were a few funnies observed during the tests where things went very strange indeed with latencies of multiple seconds for a few hundred seconds. These incidents were few and probably were the result of routing loops caused by link failures. The tests reported at NANOG were over a single IP hop between just two routers (though there were multiple ATM hops underneath). More recent tests have been performed over multiple IP hops (that is, through more than two routers) with comparable results.

What do these tests mean? For one thing, at least on an ISP backbone, IP networks are already easily reliable enough for interactive voice traffic without any QoS mechanisms. These tests did not include customer tail circuits or customer networks, which can often be overloaded, so they are not of Internet end-to-end connections.

But the test results indicate that customers with uncongested ISP links to an overprovisioned ISP (most of the big ones) will get very high-quality voice transport without having to pay extra for QoS. They may get hit for a while when an ISP link fails, but many people may put up with 0.01% downtime for a no-extra-cost service. This portends quite well for intersite, IP-based PBX connections, but is not good news for die-hard "the Internet needs circuits" folk.

Disclaimer: Whatever Harvard is, it is not a no-extra-cost service and the above observation is my own.

All contents copyright 1995-2002 Network World, Inc.