eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
It has been weeks since Comcast announced that it would be moving its “capacity management” from a protocol-oriented scheme to one based on aggregate use of bandwidth.
The protocol/application approach was found by the FCC to violate rules of network neutrality. Comcast is appealing the ruling just to protect their rights and I think they have a valid point, but that’s probably just of historical interest. After examining their new plans, it’s hard to see they would prefer the old one.
Here’s the short version of how it works: If, in any one physical section of the network, the overall traffic reaches a certain level, then they look at individual users. At such a time, if a user is using more than a certain percentage of either upstream or downstream data capacity then that user’s traffic is shifted to a lower priority status. If the user’s utilization drops below the threshold level for a sufficient period of time then they are re-prioritized.
It’s generally known that cable modem is a shared service; you’re on the same network segment as your neighbors, and it’s on these segments that the bandwidth hogging becomes an issue. How many users are on each of these segments? The coax cable on your street is all shared. At some point, it connects to an “optical node” where it bridges into a fiber network. The fiber feed through termination hubs to a CMTS (Cable Modem Termination System).
Each CMTS has multiple ports, both upstream and downstream, and these seem to be the key point at which capacity can be constrained. On average, about 275 cable modems share the same downstream port and about 100 cable modems share the same upstream port. So it may or may not be the case that you are sharing bandwidth with your immediate physical neighbors.
Comcast has approximately 3,300 CMTSs deployed throughout their network, serving 14.4 million customers, for an average of about 4,364 customers per CMTS. The bandwidth into each of these is considerable, but so is the demand for it. The CMTSs connect further upstream into Comcast’s RNRs (Regional Network Routers), and it is near these points that they are implementing the new rules, although the new equipment will work with individual ports.
Before de-prioritizing anyone’s traffic, the system looks at the traffic level at individual ports. First they determine if this port is in what they call “Near Congestion State”-a state at which performance could degrade for all users on the port. The levels, based on experimentation in their test markets, are 70 percent utilization of their provisioned bandwidth on the upstream ports and 80 percent on the downstream ports (over a 15 minute period). They say they expect such levels to be reached “for relatively small portions of the day, if at all, though there is no way to forecast what will be the busiest time on a particular port on a particular day.”
If these levels are reached then they look at the traffic utilization of users on that port. (“Users” in this case really means “cable modems.”) The “user consumption threshold” they look for is 70 percent of either upstream or downstream bandwidth, also over a 15 minute period. (All of these levels might change over time as Comcast tunes the system.)
Slowed, Not Stopped
After a conversion being performed on cable modems now, Comcast will use two QoS performance levels for traffic. The default is PBE (Priority Best-Effort) and there is a lesser BE (Best-Effort). If a user is identified as being above their user consumption threshold while their port is at a near consumption state, that user’s traffic will shift from PBE to BE. It won’t be stopped, it will be slowed down, and therefore consume less of the overall bandwidth on that port. The user will be returned to PBE state when their consumption moves below 50 percent utilization for approximately 15 minutes.
How bad can BE be? The CMTS has a scheduler, and PBE traffic gets (you guessed it) priority. BE traffic gets processed on a “space-available” basis. It’s probably not a huge difference, even at a busy time, but a heavy user will notice the difference. Comcast says they tested for the worst-case scenario, that a CMTS is 100 percent utilized and BE traffic is completely stalled, but it just didn’t happen. They claim that they have never received a complaint in their test markets for these techniques from a user that can be traced to the new management practices.
Not long ago Comcast also announced that they would be limiting monthly residential data usage to 250GB. This is not part of the same policy, although in a way it’s the same thing, but with a longer time window, and the limit is hard rather than a throttled limit. I think of the 250GB limit a different form of capacity management. (Incidentally, this works out to about 100K bytes/second all month.)
I’ve never gotten all that excited about net neutrality. It all seems kind of paranoid to me. But if you really are worried about major telecom providers treating applications and protocols and services in a discriminatory fashion then I don’t think you can have a problem with Comcast’s plan, at least as they have described it here. You can’t deny them the capability to manage their network so that capacity is maintained for all users.
Some will complain that users are being punished for using the service to its advertised capacity, and there’s some rhetorical truth to that. But cable modem has more of a shared bandwidth component than other broadband offerings and it’s just plain common sense that you can’t hog it. And the few people who are abusive are the ones who know better. If you absolutely must have huge amounts of bandwidth at all times, go buy yourself a dedicated line. Oh, you don’t want to spend all that money? Guess it wasn’t all that important to begin with.
Security Center Editor Larry Seltzer has worked in and written about the computer industry since 1983.
For insights on security coverage around the Web, take a look at eWEEK.com Security Center Editor Larry Seltzer’s blog Cheap Hack.