September 2000

Aloha! I finally had time to catch up on my reading, and was particularly taken by your article on cable modems and bandwidth in the June Cc: column of Communications Solutions�.

One of the few advantages we have out here is an (almost) overabundance of bandwidth, and I was one of the first subscribers to a local service provider's cable network service in late 1996. With nearly four years behind me, I have found -- or made -- nearly every possible mistake on broadband, and would like to share a few of my own findings.

First, my service provider was up-front about how the amount of bandwidth we would each have would be a direct function of how many people were simultaneously connected to the same node. You obviously had a slightly more difficult time finding this out....

I'm lucky in that my home (and attached office) is located out in the boonies, and there are still only four other connections to my node. The others are seldom on line during my work day, and I seldom have to share any of the 27 Mbps bandwidth...or what would be 27 Mbps if my service provider hadn't set up only 10 Mbps cable modems. Like you, I try to optimize and maximize my throughput, but was never able to quite get past the 10 Mbps barrier. I'll come back to that.

My problem for the first three years was losing my IP address on a totally random basis. This drove both me and my service provider's Level 2 and 3 support people absolutely crazy! I have 22 computers connected to my network, with the first network hub connected directly to RoadRunner. My office is quite virtual, so there were many times when no one else was here, and I had only one computer turned on. Oceanic prefers to make their profits by selling multiple IP addresses at $10 a pop I wasn't particularly interested, and that's what everyone was blaming. But when the Level 3 guy spent a day with me, he found that it was actually the gateway that was abnormally terminating!

Several months and several million dollars later, Oceanic finally decided that they really did have a bug, and it took Cisco, Motorola, and four other companies nearly another year to find and resolve that problem. Too bad they wouldn't pay me a consulting fee.

By that time, I had added a front-end router and I had started working closely with 3M's Volition fiber-optic system, which turned out to be a real windfall. 3M's local techs wired up my office with a Volition 4000 switch, fiber NICs in half my machines, and several eight-port hubs. With a 4.8 Gbps backbone, it's been absolutely astonishing how fast I can move my huge video and AutoCAD files around the office!

But I still have never been able to get by that 10 Mbps cable network barrier.

The Level 3 guy and I had become friends by then, and he swung by to listen to my complaint. It turns out that all of the Motorola cable-modems in Hawaii are maxed out at 10 Mbps!

By the way, I've been running my own tests on bandwidth and download speed, but you might want to try http://computingcentral.msn.com/topics/bandwidth/speedtest.asp if you haven't already found it.

The latency on both my network and the cable network has decreased dramatically with the replacement of half my copper network with fiber. Downloading the same file to two different computers, one with a fiber NIC and the other with copper, shows completely different results. The switch alone has dropped the packet loss to zero, but that maximum 10 Mbps sure looks different on those two machines! I've been using Download Accelerator for over a year. I don't see any difference on the fiber machines, but there's a 60-225 percent difference on the copper NICs.

My next client is due in soon, but I wanted to say that I enjoy your columns. Keep up the good work!

Ken Goldstein, Ph.D.
Exec. VP, Computer-Aided
Technologies International

[ Return To The September 2000 Table Of Contents ]