If Jerry Seinfeld did a monologue about cable modems, it might sound like this: Cable modems? What is the deal with them? I mean, you hear about them all the time, but only 50,000 were in trials last year. And where do they go, on top of my TV? In my computer? What is the deal?

And throughput--don't even get me started. This company says 30M bps. That one says 1.2M bps. This one uses two-way cable networks. That one uses telephone services for upstream data. What is the deal?

Robert Cruickshank, director of digital network technology at Cable Television Laboratories Inc., the cable TV industry's research arm, along with Jim Jordan of Netcom Systems and Bob Mandeville of European Network Laboratories, have developed much-needed benchmarks for testing and rating cable modem performance.

Under Cruickshank's direction, CableLabs, in Louisville, Colo., set out to extend benchmarks--standard yardsticks used to rate hardware performance--to the realm of cable throughput last fall.

When examining cable throughput, users should consider whether the bandwidth is shared or switched, what type of hybrid fiber-coaxial cable drops into your home or office, and what kind of processor/memory configuration a client has, said Cruickshank.

A cable modem benchmark will enable users to test modems (or obtain tests of modems) that adhere to a single standard, so true comparisons are possible.

CableLabs developed custom benchmark software on Netcom's platform. Netcom turned around and "productized it," said Cruickshank. Benchmark guru Mandeville is now using the product to rate cable modem performance at European Network Laboratories.

"We found the cable modems out there can support 50 to 100 heavy Internet browsers. Performance degradation is an issue, but the new, standard-based modems will perform better," said Cruickshank. "Cable modems are highly scalable, and our numbers will show that."

A recently agreed upon standard, joined with users' clamor for bandwidth, means the market for cable modems is about to hit top speed. All the more reason to put an odometer in place so users know what they're getting, said Cruickshank.

"Cable modems are awesome. They really are," said Cruickshank. "But we're moving beyond the trial stage, and some kind of standard in determining throughput is needed."

How awesome are they?

To understand the confusion surrounding what throughput an end user gets from different cable modems, consider some of the early entries into the market and their advertised performance.

Will a user get 30M bps of data with a cable modem? No. Some of these devices are meant to be shared among a workgroup. That 30M bps could be divided into pieces of less than 1M bps per user.

And what of the upstream rate? It appears that the Bay Networks Inc. LCP LAN City modem provides the highest upstream bandwidth among the bunch, a statistic important to consider for symmetric tasks such as videoconferencing.

But 80 percent to 90 percent of cable lines in the United States only provide downstream data--that is, they transmit data downstream at high rates. Creating a two-way cable network infrastructure is expensive to wire, and as a result, few are in place today.

But that's not surprising, because cable companies weren't thinking about telecommuting and corporate intranets when cable networks went up. They were thinking about HBO and MTV.

Because of these variables, Cruickshank said, benchmarking must be implemented so that a manager of IS knows just how much bandwidth he or she is doling out. And the service provider needs to know how much of its pipeline to the Internet it's giving out. Otherwise, the lack of understanding will mean sluggish adoption of cable for Internet access.

Not being able to deliver promised access rates is a serious threat to widespread deployment. Cruickshank draws a comparison between this rollout and the process by which 56K-bps modems came into existence.

Learning from history

Touted since September, 56K-bps modems didn't make it to market until March, and even then, only client hardware was available. Few ISPs (Internet service providers) had implemented the technology.

In the interim, users were bombarded with advertising touting the high-speed dial-up boards and their ability to connect "nearly twice as fast" as current modems.

As it turned out, real-world implementation proved to provide about a 50 percent to 70 percent improvement over 33.6K-bps modems. Good, but not what was promised. Not only were some users turned off, but ISPs flinched at spending money on the upgrade, especially since no standard was in place. This has further slowed adoption.

If users and ISPs become confused about what they're getting with cable, Cruickshank said, they will similarly shy away from the technology. And in the broadband access market, competing technologies will be available to attract hesitant customers.

"Cable wants to get into the enterprise," Cruickshank said. "But you have to look at what pipes pass by [offices]. Some have cable going by, but most have copper."

For enterprise users to adopt cable, there will have to be a compelling reason. Without benchmarks, that deployment won't happen, said Cruickshank.

Although several major vendors, including HP and IBM, recently dropped their cable modem programs, proponents of cable are not discouraged.

With benchmarks in place, an important step has been taken toward cable-based Internet access. "People who try cable modems will beg their companies not to take them away," Cruickshank said.

Source: PC Week