DSL Prime, DOCSIS Report, Fiber News, A Wireless Cloud

The Adtran CTO and everyone on the technical side has expected this since John Cioffi's FNF speech in 2005, and now ADTRAN, AT&T, Verizon and others are delivering the improved performance in the field.

The exaflood fears have proven to be garbage so far, with typical speeds rising and congestion becoming less common. That shouldn't be surprising: the sources were almost all paid by the Bells, and they ignored the Moore's Law improvements in network costs. Growth in bandwidth required per user is flat to down, but the 30-35% growth rates typical will eventually clog any network without easy upgrades. From the DSLAM of CMTS back, problems should be rare, because the routers can be replaced with far more powerful ones for about the cost of depreciation. DSL is not shared in the local loop, so the limit is the capacity of the copper to each home. Wireless and cable are shared, and he sees problems ahead.

The first step is to project likely demand for bandwidth, which he bases on the respected work of Cisco and Odlyzko/MINTS. From that data, Schneider believes download averages will range from 50 kilobits/second/home in 2007 to 150K in 2012. He multiplies that for unpredictable usage to recommend 350 kbps per household in 2009, 700K in 2012, and 1400K in 2015. These are higher than several network managers tell me is sufficient to avoid congestion problems, so I'd place those demand figures as 2011, ?2014, and 2017-2018.

With either set of assumptions, WiMAX and LTE cannot meet those targets by the middle of the next decade. Put another way, wireless in any likely deployment will be unable to deliver the bandwidth required if many homes used it for TV over the net. That's a good thing for DSL and cable, because otherwise there would be a massive shift to LTE after mass deployment ramps around 2012. If most people actually do shift their TV viewing to the net, they will still need a landline.

160/35 DOCSIS 3.0 also would fall behind mid decade by Kevin's calculations, but I'm reporting DOCSIS speeds are ready to increase rapidly. Comcast's Tony Werner tells me 120 meg DOCSIS upstream will go into trials late this year and he's hoping to go to wide deployments early next. 320 down modem chips are already sampling at the same price, and DOCSIS 3.0 is designed to go to a gigabit. The gigabit requires about a quarter of the capacity of most cable systems, so would not be practical without switched digital video (SDV) to free up channels. The FCC has just accepted SDV. Today SDV is at about 25% of homes and most large networks are likely to be totally converted in 2011-2014. Cable will need to upgrade equipment mid-decade, but my take is they have room to do so. Cablecos that don't maintain their networks may face problems.

The ADTRAN white papers and presentations go into far more, including some ingenious schemes for higher speed DSL deployments. As we go further out, everything becomes less certain. In particular, the costs of streaming quality TV have dropped so low you can support a network on advertising in the near future. There's a potential for a step function if many people choose to get all their TV from the web soon, as Henry Blodget predicts. I'm skeptical, thinking most people will not turn off their satellite or cable video programming.