Paul Plack wrote:
One of Comcast's independent techs told me their system runs wide-open to the speed-test sites to ensure good results. I don't know if he was being honest, but I never got the same speeds in normal use that I did with the test sites!


As an engineer of a CATV Internet provider, maybe I can shed some light. Our basic speed is 2M down, and we have optional packages for 4M and 6M. When we do a speed test, it shows the actual speed transferred by our equipment and the servers on the other end. Unfortunately, many of the servers providing services like Yahoo, eBay, MSN, etc. can be loaded down and even with our somewhat humble basic 2M speed we see the effects of what's happening on the other end. These effects get worse to the user as their delivered speed is increased. Folks get 'used to' seeing some sites fly, then believe there is a 'problem' when all of the places they visit don't respond with the same snappiness. I realize some of the bigger providers have their share of delivery difficulties. Some of them over-sell their bandwidth to a degree that never allows for their customers to pull what they are provisioned, but if you are able to pull good speeds to the test sites, then the ability is there to have the same performance from any site. The Internet is a highway of robust and fragile networks that are interconnected and like a chain - is only as strong as the weakest link. Now that big pipe Internet connections to common folks like us are becoming more commonplace, we are seeing the fragile-ness of the parts that cannot keep up. I sit on a LAN with 25 synchronous MB to the Internet. My provider buys 1000 MB in Pittsburgh and Philadelphia. With 2 gigs of Internet, the provider can really see the bottlenecks. You didn't even know they existed when you were on dial-up....

I'm not saying Comcast doesn't have their problems, but with multi-meg Internet connections available to the average consumer, you are going to see the weaker links; and it doesn't matter who your provider is.

Kevin Custer


Reply via email to