Fast Net News, DSL Prime, DOCSIS Report

Cablevision told me that their 50 meg boost service "always" delivers 50 meg. "So there's never a problem with congestion?" I asked to make sure I understood. "No, we have DOCSIS 3." I was on the phone to check prices for my article  on the reduced cost of 50 meg DOCSIS 3. Since the carriers have been claiming in D.C. they give clear information on speeds & services, I thought to ask about whether the connection ever slowed down.

   The answer isn't true, of course; cable data is shared and heavily over-subscribed, so there's always a possibility of congestion slowdowns. A typical DOCSIS 3 setup in the U.S. is 160 meg downstream, shared among 300 homes. If even 4 homes requested 50 meg at the same time, their speeds would have to be reduced. In practice, that happens remarkably rarely. Nearly nothing on the web except file downloading runs much faster than 3-5 meg. My research suggests that that 50 meg is actually achieved 95+% of the time. I'd bet a good network achieves 50 meg 98% of the time, but no one is willing to share data or traffic models officially to confirm that.

     This is a crucial question in D.C. because the FCC is demanding that carriers tell customers what their actual speeds are likely to be.The FCC, citing bad and inappropriate data from Comscore, said customers were only getting half the speeds advertised. That's highly exagerrated, as the technical people at the FCC explained, but it made good headlines and they repeated the claim later. Based on British OFCOM/Sam Knows testing, I'd expect 80-90% of advertised speeds on a good U.S. cable network like Cablevision.

      I first researched this late in 2008, when drafts of the stimulus plan were being tailored to make sure all the money went to the telcos while pretending to be "technology neutral." Initially, they set a minimum speed of 20 meg, because the D.C. lawyers thought cable was only 10 meg. I and others pointed to the emerging 50 meg DOCSIS. The first answer was "but that's not a real speed" so I did my research and came to the conclusion DOCSIS 3 on any decent U.S. network would be 50 meg "95+% of the time." That was based  on off the record discussion of traffic models with engineers I respected and projections from the results of sharing 36 megabit cable service. I was skeptical about sharing until MIT's David Reed told me he almost always actually measured 20 meg even though the shared coax was limited to 36 for the neighborhood.

      The next telco-inspired draft required 100 meg, easily achieved on fiber. So we came back with "DOCSIS 3 is doing 100 meg in production in other countries and could here." The first thought was they would add a quality standard, perhaps a speed reached at the 90th percentile. Some of the telcos squirmed at that, however, not wanting ever to be tested on actual speeds. For other reasons, the bill punted all the decisions to Larry Strickling at NTIA and RUS, so the debate was moot.

      The Broadband Plan decided to believe that DOCSIS 3 can deliver 100 meg. That's why they knew they were safe setting a goal of 100 meg to 100 million homes for 2020. Since 100 million home can get cable modems, essentially all of which will be DOCSIS 3, it is essentially impossible for that to fail.

      Which doesn't prevent politicians from from saying the plan is "bold and ambitious."