Translation Please: Split The Node, Or Do DOCSIS?

Here’s something to chew on: What if the channel-bonding aspects of DOCSIS 3.0 could be applied to capacity, as well as speed?

That was the question posed at a recent all-day DOCSIS 3.0 meeting, hosted by the Rocky Mountain chapter of the Society of Cable Telecommunications Engineers.

The reply, and the subject of this week’s translation: Yes.

LOAD BALANCING



Let’s step back for a moment and talk about load balancing in general.

In a two-way, shared data network, such as those used by cable modems, an important part of traffic management is to keep an eye on a thing called the “symmetry ratio.” It’s a ratio of bits sent to a cable modem, to the bits received at the headend cable modem termination system (CMTS).

Put another way, the symmetry ratio is the number of bits Customer Jane pulls into her cable modem (to stream a video, or get a Web page), divided by the number of bits she sends out of her cable modem (an outgoing e-mail, her part of a voice-over-IP call.)

In the early days of high-speed Internet services — before you could go to ABC.com to stream a missed episode of Desperate Housewives, before peer-to-peer (P2P) file sharing, before Napster, back in the days of flat Web pages and e-mail — symmetry ratios hovered around 18:1. Every one bit Jane’s modem sent returned 18 bits.

Time passed. P2P services emerged. More bits were moving out of cable modems than ever before, which flattened the symmetry ratio.

As a direct result, symmetry ratios changed — dramatically — to around 2.5:1. Every 1 bit Jane’s modem sent returned 2.5 bits. That’s a big impact, in terms of load balancing.

Then came click-streaming, for lack of a better term. That thing people do, when they go to a Web site specifically to click on a link and watch a show. (Think Hulu.com.) P2P suddenly wasn’t as necessary, to get to legitimate, copyrighted video content.

Guess what happened to symmetry ratios? Bingo. They moved again, up to the 5:1 range. Every 1 bit Jane’s modem sent returned 5 bits.

So, in a dozen years, traffic symmetry went from 18:1, to 2.5:1, to around 5:1. These are trends that matter to the overall scalability of any two-way network.

DOCSIS 3.0 VS. NODE

SPLIT
On the traffic-engineering side, there’s a constant need to apply load balancing to stay on top of such changing data patterns.

Consider hypothetical XYZ Cable, based in Congestionville. Let’s say a typical 500-home node at XYZ is already at capacity, using two downstream CMTS ports on a 1-to-1 combining ratio. That means that node and three other 500-home nodes in XYZ’s serving groups were already plucked out and attached to their own transmit/receive lasers.

One option is to split the node, as described. That way, XYZ gets two additional downstream channels, essentially, to share the traffic load so that the two existing CMTS ports feeding that node now feed two nodes. Throughput-wise, the node split gets XYZ Cable 77.6 Mbps of additional downstream capacity.

And here’s where the “more neat stuff you can do with DOCSIS” comes into play: What if you could apply more bandwidth to that same congested node — and others — by bonding a few downstream channels, then squirting them in that direction?

Let’s say it’s a four-channel bond feeding out to a node serving Customer Jane, in Congestionville. Two of the channels are the two already in use; the other two are new. Note: This still assumes there are empty channels available to bond.

But it also means that the additional 77.6 Mbps becomes 155.2 Mbps of throughput, which can be applied directly to the work of load balancing — for Jane’s node and for other nodes in XYZ Cable’s plant.