The cable vs. DSL study may be flawed, but the press is reporting on it. At least the authors of this article dug a little deeper:
Study says DSL speeds can smoke cable
By John Borland and Jim Davis
Staff Writers, CNET News.com
May 17, 1999, 4:45 p.m. PT
URL: news.com
A new survey released today pitted cable modems and high speed DSL access against each other in a broadband drag race,
and found that even low-grade DSL--at least during consumers' prime-time evening hours--was as much as 12 percent faster
than cable.
But other experts cautioned that the results, while telling for the individual neighborhoods where the tests took place, doesn't
necessarily mean that digital subscriber line (DSL) technology will always beat cable modems.
The study was done by Internet consulting firm Keynote Systems, which tracked the performance of a cable modem, a DSL
access line rated for downloads of 384 kbps, and four high speed T1 lines over the course of a month.
What Keynote's consultants found tracked generally with the way the different services have been marketed. Cable modems,
which have been marketed to consumers more heavily, performed at their peak during the day, and dropped in performance at
night when home users typically use the Net. DSL systems, marketed more to businesses, performed better at night as
employee use subsides afterhours.
Using a standard set of Web pages as a benchmark, Keynote found that the Pacific Bell DSL system took an average of 3.55
seconds to download a page between the hours of 5 p.m. and 11 p.m., and 4.30 seconds during the daytime hours.
The cable modem system took an average of 3.97 seconds to download the set of benchmark pages during the evening, and
3.68 seconds during the day. That means that DSL bettered cable modem performance during the evening by about 12
percent, while during the daytime DSL connections were about 17 percent slower than a cable modem connection.
The results were obtained by measuring the download speed of a specific set of Web pages continuously over the month of
April, all from neighborhoods in the San Francisco Bay Area. Identically configured, high-end Pentium computers systems
running Microsoft's NT operating system were used in the tests.
What do the tests mean?
Experts were quick to point out that even where DSL connections bettered cable modem download speeds, variations in
performance could be attributable to each system's different architecture, rather than the superiority of one type of connection
over another.
Cable modems are based on a shared network, in which each subscriber in a given neighborhood shares access to the same
cable stream. Thus, if a number of subscribers try to access the system at the same time, it cuts into the limited amount of
bandwidth available, slowing down performance at peak hours.
DSL systems, in contrast, give each subscriber a dedicated line to the telephone company's central office. From that point on,
the signal can run into interference, however. According to industry insiders, telephone companies like Pacific Bell typically
"oversubscribe" their consumer DSL systems, which can overburden the link between their offices and the backbone network.
This means that if many users are on the system at once they can still run into traffic jams on the far side of the telco's central
office.
Keynote said their results indicated broad usage patterns for the different technologies, but shouldn't necessarily be taken as
specific benchmarks for the access systems.
"It depends on how much your neighborhood looks like our test neighborhood," said Gene Shklar, a spokesman for the
company.
Analysts generally agreed that the test mirrored usage patterns, but cautioned that the results should not be taken as universal
since they used just a single example of each technology.
"Results will really vary by geographical location and by service providers," said Michael Harris, an analyst with Kinetic
Strategies. "There are a bunch of different variables that can affect performance."
Patti Reali, telecomunications analyst with Gartner Group, takes exception to the survey's findings, too.
"I don't think that this particular analysis can be judged of all [cable] systems across the board," said Reali. For instance, there
are some issues with how TCI's (now a unit of AT&T) cable systems are built to handle data traffic, she said. "In general, many
of the complaints that have been issued [about cable modem service] have come from TCI systems."
TCI is the main provider of @Home's cable modem service in the San Francisco Bay Area.
In short, cable operators build their networks differently, and some of the discrepancies in performance between DSL and
cable modem service could evaporate if the test was conducted in different areas of the country.
However, the survey does point to an important issue when considering how AT&T, Comcast, and Time Warner will offer all
of the advanced services being promised--high speed Internet access, telephony, video on demand, and others--over a cable
network.
"Engineering and design of the network will be very important in delivering services that people will be willing to pay for," Reali
said. |
