Gigabit Ethernet Cable? - Page 2

Re: Re: Re: By the way MCL...

Look. I don't know where you get off, but an HFC network has C in it. That's what an HFC network is: Hybrid. Fiber. COAX. Don't come in here making big noise about non-coax HFC networks. There ain't no such beast. Capisce?





I've "worked with large ISPs" in the past, too. I don't care what's on the other side of the headend. Cable modem deployments in America, by the major players (AT&T, Ameritech, Comcast, Roadrunner, et alia) use RG-58,59, or 6 for the last mile. Coax from the customer premises to the headend, which is a CMTS using QPSK or QAM modulation for the data channel. And that particular configuration has a max throughput of roughly 35-45Mbps.





If someone mentions "cable modems", in a forum predominantly occupied by American consumers, and they don't provide any further context, that's a significant clue that the person making the reference expects the context to be supplied by the reader. And, that context, being supplied by an average American, is going to be the cable modem service delivered to 99% of American homes that have cable modem service, and it's going to look exactly like I've just described above.





So you've got some experience with *special-purpose, limited deployment* networks? Not only am I unimpressed, I'm thoroughly so, as you seem to think that "ATM" means "no coax" (at which point I'd suggest you go study a bit more).





You don't know me from Adam (pun intended), and your first communication with me is an attack on my knowledge and expertise. That's unwise. And, as many people can attest, publically embarassing for the attacker. When I don't know something, I readily admit it. You, on the other hand, seem to want to extend your knowledge of non-coax last-mile deployments to a network architecture that in general requires it, and in the context of this particular discussion, specifically demands it -- because the topic at hand was the throughput on the average cable modem system.





I stated that your average cable modem system is capable of delivering 35-45Mbps to the door, today. I then stated that you'd not see it, because the average cable modem system is underbuilt and oversold, intentionally. I then further stated that the reason the throughput is 35-45Mbps today is due to the modulation scheme in use at the headend. I'll readily admit that RG58,59,and/or 6 can handle higher throughput. I'll also point out that for a run of any significant length, you'd probably not want to use it anyway, due to signal attentuation and leakage at the various junctions. It'd fail utterly in an MDU at those speeds (and this you'd know if you were at all familiar with the vagaries of TCP/IP, and what a borderline signal on the wire can force the protocol stack to do).





And yet you continue to bluster on about topics that have nothing to do with anything I've stated: you've ranted and raved about ATM (as though ATM is impossible over coax). You've gone on and on about something you think exists: An HFC network without C in it. You've repeatedly insulted my intelligence.





What are you expecting, thanks?





The problem with technical topics is this: In most everyday situations, when someone's being a complete and utter idiot, it's easy for everyone to spot. When the subject at hand is a relatively obscure and/or arcane technical topic, it's more difficult for the layperson to spot.





Rather than keep my mirth to myself, I thought I'd act as a klaxon for everyone else, so we can all have a laugh at your expense. Because you have indeed removed all doubt by opening your mouth.





Now, crawl back into your little hole, and consider holding your tongue about what someone does and does not know next time. Particularly when it's you who is on the attack.





I won't bother handing you your arse. I left it in the corner over there. Fetch.


And, as to my credentials, since you want to push the matter: BS, MA, and abd, experimental cognitive psychology. Invited researcher, German Institute for Artificial Intelligence. 15 years proefssional systems and network administration and design. Vice President, SAGE Certification. Advisor, LPI certification. Author, SAGE Certification. Author, USENIX book: "Documentation for Systems Administrators". Former clients include Sony, the National Opinion Reseach Center, WebTV, the US's largest nuclear power concern, my role as CTO of a $100mil consulting firm. REPEAT invited technical expert, TechTV. And about 20 peer-reviewed scientific articles and book chapters.

Oh, and my current position, overseeing the computing and network infrastructure for one of the world's largest and most ambitious scientific undertakings.

You were saying?

^^^

Thats the reason i stoped arguing with you MCL.. u know ur stuff too damned well.

Bah with all that answer my question. Is incapable. Is capable. This is for overall North America, im sure there are many areas capable of it. I want to know if the overall internet cable modem infrastructure is capable of this. Also how soon will we start to see the ability at a consumer level for these faster services to be purchased? I mean I dl from my modem at around 400Kb/s from a good server. I know this is what most peak out at. I also download files to my remote servers in St. louis located at the bandwidth exchange building. Runs 10k ide hard drives, using 2 3com 100mb nics, hooked to cisco 3524 switches and cisco 7500 routers. with 1 gigabit fiber lines running to it. This is backed to 6 oc 192s redundent. Providers are at&t uunet, qwest, sprint, level3 & genuity. Ok so when I dl from there like lets say downloading the halo server pack from microsoft server. I hit 1000KB/s From home did 400. Now my dl speed from home is 3500, so Im guessing that dl speed is still under 10Mb/s from my remote server. So something here doesnt seem right to me. Im on a line capable of dling at 9.6Gbps so why am I still dling at a 10mb/s speed? Lucy got some splainin to do. hehe.

1) Cable modems as commonly sold in the US already SURPASS T1 speeds, for downstream. For upstream, they're capped at (depending on carrier) 128kbps, 256kbps, or 384kbps.

2) It sounds like you're confusing b and B. T1 speeds are a bit over 1.5Mbps. Mega BITS per second. That's about (with a bit of rounding), 200kBps (kiloBYTES per second). You say you average around 300kBps, so you're already above T1 speeds downstream.

3) Backhaul matters. I know cable companies love to say, "128k/256k/384kbps up should be fine", but it's simply not so. It depends upon the type of usage, which in turn depends upon the protocol in use, and/or the behavior driving the use of that protocol. Try piping a remote X11 session over your average cable modem (1500kbps down/128kbps up). You're going to see a LOT of lag. That's because X11 is chatty on the backchannel.


4) Now, to your last question, the confusion in pipe size versus throughput: There are a number of reasons you could be seeing the behavior you're seeing, but the most likely is that you've got a bottleneck somewhere. A pipe is only as fast as its slowest part. I could have quad OC192s feeding a Black Diamond Extreme in my living room, but if all I've got coming out of that BDE is a 10Mbps, 10^T connection to a switch that is in turn connected to my desktop, all I'm going to see is 10Mbps, minus protocol and frame overhead, and modulo any degredation in linespeed due to physical layer quality and/or backplane issues (don't get me started on BDE backplanes).

The long and short of cable modem speeds is this: You will get, at best, whatever the cable company wants you to get. It's the same principle airlines operate on: A plane has so many seats. They sell so many + X seats, knowing that a certain, statistically-demonstrable fraction (which is hopefully equal to X) will not show up when the flight's scheduled to depart. Most of the time, they're right (because they fudge X enough on the conservative side to ensure the flight is as full as possible without going over). Sometimes, they're wrong, and they have to ask people to surrender their seat.

Same things true in cable modem land (moreso than you might think, given the "bucket" mentality used to apportion bandwidth at the CMTS): The cableco sells X number of connections per headend, knowing/hoping that a certain percentage of people will never exceed X bits per unit of time, and that a certain percentage will not be using their connection at all a certain percentage of the time.

Sometimes they're right. Sometimes they're wrong. When they're wrong, you see packet loss and latency spikes. When they're right, you don't notice anything amiss.

The problem is that they oversell to the point that they've got nothing held in reserve to accomodate those times when they're wrong. That's WHY you see the latency spikes and the packet loss. But they're not driven by technical competence; they're driven, as I mentioned in my original post, by the old "Ma Bell" mentality: Meter all usage, and find profit in gross load, rather than net packets. It's a cell-switched v. packet-switched mentality, and the fact that it's a game of margins already highlights the fact that thinking in a cell-switched manner's simply silly for these economies of scale.

Unfortunately, we can expect to see the cellphone carriers realize this before the cable modem providers (due to larger penetration, thus smaller margins), and they still haven't tumbled to it.


The whole point of this is: There are real-world performance capabilities, theoretical maximum capabilities, and services designed to maximize income while minimizing outflow. Cable companies are cognizant of the first, will never approach the second, but design and deploy around the third. And, unfortunately, the realm in which profit exists both affects the first, and abuts its boundaries. Which is why performance, on the whole, tends to suck during peak hours.

(Note to the technically competent (which doesn't include my recent antagonist): This doesn't even touch on issues of backbone traffic draining and other esoteric routing issues that can similarly affect performance, but given the growth of the largest three cable co. networks, I can tell you that they prefer a "drain early, drain often" approach that is good for short-haul (local) traffic, but really shows strain on coastal and intracontinental latency. And I won't even mention the severe breakage that at least two of the top cableco's have exhibited w.r.t. RFC1918. Suffice to say they're advertizing routes they Just. Shouldn't. Be.)

Ok Mcl You really do know your stuff, but I mean kind of a over answer. Let me do it this way. For my first question

Yes the current cable modem infrastructure

1. Is incapable of attaining speeds over 10Mb/s

2. Is capable of attaining speeds over 10Mb/s


A simple 1, or 2 would sufice. And the questions had nothing to do with my cable modem. I have 3500 dl, and 384 up. Im on comcast, and have no problems. I was speaking of my remote server which is fully capable to download at 100Mbps. Yes I know the diff with b, and B. My point was, even the fastest servers out there dont push anywhere near to 100mbps. So the advantage at this time of having the capability is not much. Its like having a 100 gallon bucket, and only 10 gallons of water to fill it. I hope within 5 years we are all pushing near 100mbps speeds, the technology is surely there, its just the infrastructure. The companies are probably mostly putting off the upgrading till their equipment gets older. I mean its a cost vs profit thing. Will they make more money if they upgrade, and offer the customer more bandwidth at the same cost? no probably not. Well I dont want to go on for pages, so Mcl please just put a 1, or a 2. t hat will sufice. hehe

2. With the caveat that certain older cable modems only have a 10^T Ethernet port, so the BEST they could deliver is 10Mbps.

And with the understanding that you won't see anything over 10Mbps, let alone near it, any time in the near future.

Oh if anyone knows of a download link that pushes over 1000KBps let me know. Id like to run a test on it.

Hahaha, hopefully by the time I'm done school in 5-7 years, they'll have some nice juicy/pricey pipes for me to chew on .

Well, there're several Internet2 sites I could mention but, well, you'd need an Internet2 connection.

http://download.microsoft.com/downlo...TrialSetup.exe That hit 1000KB/s when I dl'd it from my remote server.

I should say I did that at non peak times around 9 am this morning. People may be dl'ing the demo more, and slowing the network.

Its all upto your local office. If they are lax in upgrades your service will be bad. You also need to account for the fact that it could be your buddies computer.

Dont you guys have DSL available? Here in Ontario, everyone who is speed-concious have replaced cable a long time ago. Sure, cable is fater IF IT WORKS. But most of the time, the neighbourhood is so clogged up, cable bluntly sucks. I have a 1.5Mbps down (clean; 1.7 raw) and about 300kbps up. I also can get an 'ultra' service which is 3Mbps down.

Lets not start the cable vs. dsl war. lol Dsl loses. But it is once again a provider issue. If you have a crappy cable provider, and a good dsl provider dsl will win. But dsl is bottlenecked plain and simple. The potential with cable modems is higher than any dsl. Now as far as the peak time argument. I dont ever see that. I did two years ago, and after about literally 70 phone calls I got to admit that their upstreams were not balancing on the node. It was a firmware issue. They hadnt oversold the node, but there would be 5000 customers on upstream 4, and 300 on upstream 5. So if you got stuck on 4 you were fubared. I found a trick around it however, id get into the config on my motorola, reset to defaults, then make it restart the modem. Id keep doing it till my upstream switched channels. They've since due to my complaining upgraded that equipment. And Ive had 1 15 minute outage in the last 8 months. I always test right around 3200 dl, and 368 up. Im set up on my account for 3500 dl, and 384 up. So im pretty much dead on where Im supposed to be 97% of the time, regardless of peak times ect. Cable wins as long as you have a good provider everytime.

heh, i just hit 1.5MB/sec from that halo link...and from the EA ftp server i usually get around 3-3.5 MB/sec here at RIT

Guys, guys, guys..... All I was saying is that if cable sucks in his area (like in mine), he should look into DSL (if its available).

I never downloaded much music in the first place. Generally, I like things nice and speedy. Anything I need, I get it quick.