Cable’s upstream bandwidth capacity has gotten a lot of attention recently, and with good reason. There’s no imminent crisis for the industry, but even in the residential market, cable operators have to start planning now for demand increases over the next decade. On that point, Motorola Senior Manager of Systems Engineering Dean Stoneback recently presented a paper modeling out likely demand growth through 2020 and illustrating the supporting architecture required.
The graph above shows different network configurations and plots varying demand growth rates across a number of years. The bottom line in purple maps to a cable plant using the spectrum between 5 MHz and 42 MHz for upstream traffic with 250 households passed per node. At the lowest projected levels of 30% growth per year, the network runs into trouble right at about the decade mark. At the highest projected levels of 60% growth per year, that window of acceptable performance drops in half to about five years.
There are a lot of assumptions that go into the graph above that are difficult to detail in a blog post. (Not to mention highly technical for this forum.) However, for context, we can assume that in 2010 we are starting with peak upstream throughput per user of 5 Mbps, a 50% penetration level, 2% concurrency, and an initial node size of 500 homes passed. (See the Motorola bandwidth calculator for details on how these variables impact required capacity.) Over time, peak upstream demand is assumed to grow both because of higher-bandwidth applications and because of higher concurrency levels. To counter that, it’s assumed operators will continue to decrease node sizes in the coming years, bringing service group numbers down from 500 homes passed down to 250 or 125.
Whatever growth rate actually occurs in the next decade, it’s interesting to note that it took four to five years from development to deployment in the rollout of DOCSIS 3.0. That means cable operators need to start planning now for upgrades to address upstream capacity – even if the consumer use case takes several more years to prove itself out.