Blandwidth

The graphic on the left is from a Vonage test of the connection at a friend’s house near Boston. Comcast cable is her provider. The test was on her computer, which is connected directly to the cable modem. I thought that test result was exceedingly lopsided and Old Skool in respect to upstream performance, so I conducted a different test on the same connection with the same computer. The result: 11958Kbps down and 358Kbps up.

Comcast can do better than that. I suspect the only reason they’re not is because they’d need to “bind” some number of channels that would otherwise carry television. Whatever’s going on, it’s clear that the Net is just gravy on TV. Feh.



9 responses to “Blandwidth”

  1. Comcast customer here. Vonage shows me with a 10.0Mbps download and a 1.95Mbps upload speed.

    7,769Kbps download with Speedtest.net and 939Kbps upload.

    11.2Mps down and 2.24Mps up Vonage after I turned utorrent off. Oopsie!

    10,014 down and 1,041 up with Speedtest.net.

    And I’m going through a router. Still an order of magnitude than your friend is getting on the up side.

  2. They also “cheat,” although they are very transparent about it — Comcast’s (somewhat) new PowerBoost(tm) gives you something like double the bandwidth for the first 5-10MB of downloads. It’s actually fairly nice, although a bit odd as a feature. It definitely skews the bandwidth test results, though. If you pay them another $10/month you can get up to 768kb up instead of 384kb. Peanuts, I know.

  3. Back home now. Over wi-fi, 6Mbps down, 1Mbps up. Steady. Over Cox Cable. Possibly the best of a weird breed.

  4. Hey Doc, here are the results for my Cox cable connection using Speakeasy.net/speedtest.

    13290 kbs (1661.3 KB per sec.) Download
    1145 kbps (143.1 KB per sec.) Upload

  5. It’s a question of network engineering. Cable Internet has two channels, one for downloads and the other for uploads. On the download side, there is only one transmitter, so he can jump in the cable immediately each time he wants to transmit, using addresses in the packets he sends to determine who gets the stuff. Every receiver sees every download packet. The cable is running at something like 25 Mb/s (more or less) and that’s pretty close to the utilization.

    On the upstream side things are complicated by the fact that you have multiple potential transmitters on the same channel, so they need a control procedure to ensure that they don’t step on each other’s packets. The overhead in this procedure – some sort of time slotted thing overlaid on a polling system – keeps the upload speed down because it drops the cable throughput to something like 25% of the raw signaling rate. It’s kinda like WiFi, where you have systems with signaling rates of 54 Mb/s that can only achieve throughput of 26 Mb/s because of network overhead. If the cable system wanted to offer higher upload speeds, they’d need to set aside more channels or deploy more headends and populate each cable less densely. But their main revenue source is TV, not Internet access, so you’ll have to put up with sharing your cable with lots of other people until that changes.

    You’re ahead of the curve, so you get the arrows in the back.

  6. Thanks, Richard. Helpful stuff.

    I’ve talked with Cox about this, and they explained it in terms of “channel bonding” (I think that’s it, or something close). The problem with provisioning there is a kind of chicken/egg issue. In the past they said there was no demand for upload capacity, but they also didn’t provide much of it, so how could they tell? That said, they’ve increased it, and for that I’m grateful. I moved my laptop from wifi to ethernet, connected straight to the router, and I’m getting close to 7Mbps down and better than 1.1Mbps up.

    Now here’s the thing. If Comcast provided that, they’d be somewhat competitive at the house we’re renting near Boston for the next year. But that slow upload speed is a downer. Some asymmetry is cool, and totally part of the program with cable. But >10x difference is a bit much.

    Meanwhile, the other choices are RCN and Verizon FiOS, both fiber-based. Not sure if it goes to the premise, but it seems to get to the large new boxes hanging from the poles outside the house. RCN is 20/2Mbps, while Verizon is 45/5.

    I would think that, in a location like this one, thick with techies and academics, somebody (RCN, I would think) could break out and stress the Net part of “triple play”. But so far they’re not. So we’re going with Verizon.

    Interesting history… There were two big reasons we moved to Santa Barbara in ’01: 1) much better bandwidth than we were getting in Woodside (where we suffered with “IDSL” that barely beat dial-up), and 2) Peets Coffee. At the time Cox was producing speeds of up to 7Mbps on the downstream (I forget the upstream figures) in Santa Barbara. That almost locked it. Then it turned out there was a Peet’s on upper State Street. That nailed it.

    Now, while we’d be hanging for the next year near Boston in any case, having truly competitive bandwidth providers (no duopoly here, and it shows) was a big reason why we chose the house that we did.

    The market at work, no?

  7. When the cable Internet specs (it’s called DOCSIS, BTW) were written, the expectation of cable providers was that downloads would dominate uploads by at least 100:1, that being the conventional distribution of interactive applications like web browsing. So they were pretty radical to break out as much as a 10:1 ratio. They must have been expecting a lot of e-mail. Now the market has changed and with it the engineering will have to change.

    Verizon’s FiOS is by far the coolest thing going in broadband today, and I have lust in my heart for it. I’m on Comcast with not even a DSL option, and the monopoly really stinks. I used to have a sentimental attachment to cable Internet because I wrote the software the @Home installers used, but that was long ago and far away.

    Enjoy your Verizon, you lucky dog.

  8. The whole upload process is hampered by the dynamic allocation of IP addresses used. In the days of dial up, the ISP needed fewer IP addressed when they were dynamically allocated, but this is no longer true. I already pay for bandwidth, why should I pay more to allow others to pull data from my system instead of my pushing it?

  9. […] Doc Searls Weblog Blandwidth Posted by root 3 days ago (http://blogs.law.harvard.edu) Cable internet has two channels one for downloads and the other for uploads when the cable internet specs it called docsis btw were written the expectation of cable providers was that downloads would website your comment powered by wordpress and tarski ho Discuss  |  Bury |  News | Doc Searls Weblog Blandwidth […]

Leave a Reply

Your email address will not be published. Required fields are marked *