Please create an account to participate in the Slashdot moderation system


Forgot your password?
DEAL: For $25 - Add A Second Phone Number To Your Smartphone for life! Use promo code SLASHDOT25. Also, Slashdot's Facebook page has a chat bot now. Message it for stories and more. Check out the new SourceForge HTML5 internet speed test! ×

Comment Re:Don't worry, it'll get "better" (Score 1) 317

To be completely honest with you I am not sure which bandwidth capping packages will be offered. All I got was a memo from a higher up saying that it was only experimental, and there where not any plans to spread it beyond the testing in Texas. If it does not go over well with customers, then they wont be continuing with it I am sure.

VOD is a different beast altogether. It does not use the high speed data infrastructure at all, they use their own QAM's (MQAM's soon when 'start over' rolls out) which only ties up more frequencies on the coax. Right now with DOCSIS 1.1 we are only using 2 32Mghz bands on the coax for HSD, one for the upstream and one for the downstream. I can only speak for my division here in Bakersfield, but I know that as of 2011 we are planning on dropping our analog signal from the plant altogether, freeing up a lot of frequencies on the cable, making plenty of room for DOCSIS 3.0, which one modem can "talk" to multiple downstream giving more capacity.

We are constantly doing node splits on the fiber transport to lower the number of homes passed, putting the new node on separate upstream/downstream. We also do a lot of moving around of the nodes themselves. Typically we like to keep the modem counts around 400 per downstream, this depends largely on what bandwidth packages are offered in your area and what usage is like. Most of the time a decombine (moving a node to a new downstream) happens when we traffic hits certain thresholds on the downstream. Each upstream we keep at around 100 modems (4 upstream share the downstream port). This seems to support our area pretty well, we are getting ready to change all of our standard package speeds to 10 down 1 up and make the premium speed 15 down and 2 up. So to answer your question, yes most of the congestion issue occur in the last mile access. Bakersfield is handled by 2 OC48's to our LA POP sites, one to L3 and one to B-Bone. Right now it's 7:00 PM and bordering on our peak time. Here is the current traffic across one of our OC48s.

    30 second input rate 1080735000 bits/sec, 168436 packets/sec
    30 second output rate 340324000 bits/sec, 144683 packets/sec

As you can see we are not even using half of the 2.4Gb/s the link is capable of. The majority of traffic shows as incoming here because it is the last hop before it get routed back to the modems, input is the modems downstream traffic. I just did a check and there are close 60,000 modems online right now. Here is the downstream stats from one of the CMTS with approx 400 modems on the downstream.

    5 minute output rate 20707000 bits/sec, 3834 packets/sec

Just over half of the downstreams 37,500 Kb/s. Mind you this is sharing bandwidth with the MTA's for VoIP traffic as well. Though the issue is the last mile access, the nature of 90% of our users traffic (and most peoples for that matter) does not make it that big of a deal. Most peoples downstream traffic happens in bursts and sustained upstream traffic is even farther between.

Being on the front line I can tell you just how busy we really are, keeping everyone happy and surfing along at the best speeds possible. The cost of one 5x20 cable card for a Cisco 10k chassis is around $100,000. That's 5 downstreams and 20 upstreams. I have added 2 of those cards to accomodate growth and releave congestion in the last 3-4 months, and those are just in our 2 fastest growing hubs.

If you're still with me I know that was long and probably riddle with typos, but I hope it sheds a little on the subject.

Slashdot Top Deals

A list is only as strong as its weakest link. -- Don Knuth