Cable's Upstream Gap

5:30 PM -- Verizon Communications Inc. (NYSE: VZ) booted up its Quantum FiOS Internet tiers Monday, and much attention was heaped on the new 300Mbit/s downstream tier that essentially doubled the speed of its previous high-end offering. (See FiOS Speeds & Prices Take a Quantum Leap .)

While 300Mbit/s is a nice, round number that gives Verizon's marketers something to toot their horns about, it's also something that cable's in a position to counter. Today's Docsis 3.0 modems can bond eight downstream channels and produce the kind of speed bursts that Verizon now has. Intel Corp. (Nasdaq: INTC)'s new Puma6 D3 chip and its 24-channel capabilities will put cable within shouting distance of 1Gbit/s, and the 32-channel wideband chip believed to be on Broadcom Corp. (Nasdaq: BRCM)'s roadmap will blow right past that. (See Intel's New Docsis 3.0 Chip Guns for 1-Gig and Broadcom's Next D3 Chip Will Leapfrog Intel .)

So, cable's got the goods in the downstream. But, it's a matter of whether the operators even have that many channels to spare, and if it makes any sense to use them just to create a tier to toss into the marketing hoopla.

Cable doesn't, however, appear to have a quick answer to match what Verizon FiOS is now doing in the upstream, offering 65Mbit/s for its fastest FiOS Quantum tier, and 35Mbit/s and 25Mbit/s for the other two. I get the sense that Verizon could go higher than that if needed, but those speeds already do a decent job of exposing cable's historic high-speed weakness -- its rail-thin upstream path.

Among some of FiOS's big cable competitors, Comcast Corp. (Nasdaq: CMCSA, CMCSK)'s fastest residential D3 tier offers upstream bursts of 20Mbit/s, which isn't too shabby. Cablevision Systems Corp. (NYSE: CVC)'s comes in at 15Mbit/s, which isn't bad either. But Time Warner Cable Inc. (NYSE: TWC) is trailing them both with top upstream speeds of just 5Mbit/s.

Intel claims its new D3 chip can hit 320Mbit/s by bonding eight upstream channels, but that comes with the assumption that cable's got eight clean upstream channels with which to work with in that 5MHz to 42MHz range. It's increasingly likely that cable will continue to bond upstream channels early on to boost the quality of its existing upstream speeds rather than shoot for higher and higher numbers. And, if history repeats itself, cable's upstream speeds will continue to lag behind the downstream, anyway.

I'm not the only one who recognizes that cable's got to establish its long-term vision for the upstream. The cable guys see it too. There's still a lot of talk (but little action) on a mid-split that would expand cable's upstream to 85MHz. And there's discussion under way on Docsis 3.1, a purported enhancement to the specs that will look at higher frequency modulations and perhaps raise the upstream ceiling to 200MHz to 300MHz -- enough to put cable on the road toward a 1Gbit/s upstream and solve some of those long-term issues. (See The Docsis Addendum .)

But mid-splits can't be done overnight, and the new spec appears to be in its infant stages. So that tells me that we're still years away from seeing anything truly shiny and new when it comes to cable's upstream. (See The Docsis Addendum .)

And that may be okay. Cable has managed to hold its own against FiOS and other high-speed competitors, most recently grabbing 75 percent of the net broadband additions in the first quarter of 2012, according to Leichtman Research Group Inc. (LRG) 's study of top U.S. broadband ISPs. It's still gaining ground overall.

So, do big speeds really matter? I buy into Verizon's argument that it will in the downstream as consumers use more and more connected devices to stream video at the same time off the home network. But cable, despite the big gap on the upstream, may have enough time to fix that part of the problem before it really starts to hurt.

— Jeff Baumgartner, Site Editor, Light Reading Cable

rarpsl 12/5/2012 | 5:29:47 PM
re: Cable's Upstream Gap

As your download speed increases, your upload speed must increase also just to allow the full download speed. To be able to download at full speed you need to use your upload to send "I received the download packets" ACKs. As your download speed increases, the number of ACKs increase and so does the delay for the sender to receive the ACK of the first packet sent. Unless the ACK is received and processed in time to send another download packet you will be throttled to a slower speed.  The basic idea is that your download speed is constrained by the need to request more packets via the ACKs of the packets you have received. Thus you need enough upload speed to get the ACKs to the sender with enough time to triggger the download packets to keep the flow going at full speed.

macemoneta 12/5/2012 | 5:29:47 PM
re: Cable's Upstream Gap

Whether the upstream matters depends upon whether you buy into cloud services for consumers.  In order to provide transparent replacement of local functionality with cloud based services, the upstream matters.  Google has recognized this, and it's the reason behind its symmetrical 1Gb/s trial deployments in Missouri, Kansas and Connecticut.   If you're not willing to use a slow 15-year old disk drive with your computer, you're not willing to use a less-than gigabit upstream for the function. 

Duh! 12/5/2012 | 5:29:46 PM
re: Cable's Upstream Gap

Assuming it's TCP you're talking about, the rule of thumb is about 8:1.  The issue is not only with keeping the window open, but also with TCP's slow start and congestion avoidance mechanisms.  See RFC 3819 for details.  There are solutions to the problem involving proxies in the Cable Modem that drop stale ACKs.  I recollect that at least some CM ASSPs (TI, if I remember correctly) implement that solution.  See RFC 3449 for details.

Jeff Baumgartner 12/5/2012 | 5:29:46 PM
re: Cable's Upstream Gap

I realize that this would more more appropriate to mention if Videotron of Canada did something wacky and totally unbelievable like invade the US and overbuild FiOS territories, but just thought it was worth mentioning something that they're doing today that offers indication as to what their cable peers south of the border could do now, if they really set their minds to it.

Videotron's latest high-end tier maxes at 200Mbit/s down and 30Mbit/s up. I costs C$209.95 unbundled, or C$199.95 when combined with at least one more service, so it's unlikely to get alot of takers. But I'm guessing the marketers sure like having that one in the arsenal. JB

Sign In