re: Car Pools & QOSI've heard a lot of talk from telcos about their plans to deliver video and TV on demand over their DSL networks.
But the backhaul from most DSLAMs is highly over-subscribed - 50:1 in the case of BT here in the UK.
Regardless of any QoS scheme the carriers put in place, there isn't enough bandwidth for video to operate in these networks. VoIP maybe, but not video - even if it is crammed into a 1Mbit/s Media 9 stream!
The reason that QoS is an element and is a problem with the TR-58/59 setup is that Video creates the DSLAM as a blocking element of the network. Not just the DSL line itself.
The question becomes how you manage the QoS through the ATM subnetwork to the customer. The claim in the article is that most DSL modems don't support 2 PVCs. Well, they also don't support using Ethernet or IP QoS. That is why for Video over DSL networks its almost always imperative to have interoperability end to end to be confirmed with the vendor set chosen. These systems (all following standards) have lots of issues with QoS management as one factor but there are many others (example: your ISP and your video provider are different).
Anybody thinking that today, any random choosing of Encoder, VoD Server, Middleware, IGMP Router, DSLAM, Modem, and Set Top are going to work is smoking dope. There are specific needs that each element has. There is no reuse of elements (unless one is lucky) from a high speed internet only setup.
That is why there is no debate. Company A solves the problem one way. Company B solves it another way. They all have their plusses and minuses, BUT they are not interchangeable.
re: Car Pools & QOSHuw, Where do you see Ethernet DSLAMs fitting into this picture? I know almost all of the existing base of DSLAMs are on ATM, but a fair chunk of new deployments use Ethernet.
re: Car Pools & QOS The basic issue is where you place IGMP multicasting points for Broadcast Video and how much unicast traffic you expect to have. If Broadcast dominates, then your need is to transport your broadcast content which can be run over a single GigE for 150 - 200 channels even today.
re: Car Pools & QOS Biggest issue is what happened to UT Starrcomm at Telmex. The current generation of IP DSLAMs are cost optimized for HSI. They fall down from a bandwidth standpoint once video is applied.
re: Car Pools & QOSWith a disclaimer that I don't know what I'm talking about, never having dealt with DSL, but here goes anyway:
I would think that the modem wouldn't have to know about IP priorities, any more than an ethernet PHY does. Rather, whatever presents traffic to it has to so in the right order and with the appropriate preference for discards. So, whatever drives the modem has to do the QOS.
Now one would think passi paru (I always use Latin when I don't know what I'm talking about) that the same would hold for the multiple PVC case,- but presumably the modem is cellifying the data and attaching the (single, canned) ATM header.
Corrections welcome!
-----------> The question becomes how you manage the QoS through the ATM subnetwork to the customer. The claim in the article is that most DSL modems don't support 2 PVCs. Well, they also don't support using Ethernet or IP QoS.
re: Car Pools & QOSI've been chatting to Graham Beniston about this column.
As some of you will know, Graham has produced a number of webinars and reports on DSL topics, for Heavy Reading, and has taken a keen interest in the DSL Forum's TR59 and its influence on the architecture of next generation broadand networks.
Graham takes an opposing view to Huw - he thinks you do need 2 ATM PVCs - and I've been trying to persuade him to post some of his arguments for thinking this on this message board. No luck so far. He wants to respond with a column of his own, and to do some research before hand.
However, one of the points he was making to me personally was that the issue of 100 million subscribers having to replace their DSL modems in order to support twin ATM PVCs might be a bit of a red herring. The chances are they'd have to buy some other gear anyhow, to be able to use these new types of service, so they might as well buy some gear that incorporated an upgraded DSL modem at the same time.
Someone else also pointed out to me that expecting subscribers to chuck out their old DSL modem isn't as wild as it might seen. Consumers regularly chuck away their cell phones.
re: Car Pools & QOSHi seven, I understand and agree with what you're saying about one carrier, one broadband architecture.
But surely there's still room for debate? How do carriers decide on these architectures in the first place? There has to be a process of discussion.
It's quite possible that the "first generation" broadband infrastructures that were built over the past few years are now in need of a good shakeup so they can begin to run premium margin applications, such as video and VoIP. In some ways this is good news as there will be a lot of new equipment required, and somebody needs to build and sell it.
But there's a bigger question too - how does the wholesale broadband provider get some revenue from premium applications that run over their network? It's the retail SPs who get that revenue.
Most regulatory regimes of which I'm aware seem to include this retail/wholesale split. (Note, can someone comment on the regulatory situation in Asia/Pac?).
In the US, Carriers own and replace the DSL Modem. Thus a sub with an existing DSL service and wanted to upgrade it with IP Video over DSL would have to replace that modem (at the carrier expense). There is an entire home wiring issue that has not been part of this discussion yet (i.e. How do we get Ethernet through the house and how do we hook up multiple TVs in different rooms).
I guess my question to you at light reading is why you have never had a webinar with folks that actually have deployed Video over DSL? Skip the equipment companies, go the IOCs that have done it. Theoretical arguments about PVCs are one thing. Getting a router that forwards traffic on 100% of its ports 100% of the time AND can switch channels in a reasonable time AND is cheap (Good number as a toy to use is 100 subs per GiGE - go price out end office router costs with that) is a non-trivial challenge.
re: Car Pools & QOS It does matter Priam. There are real time requirements on the return path of the set top boxes. If the DSL Modem is not capable of getting that traffic prioritized, customers reach unhappiness.
But the backhaul from most DSLAMs is highly over-subscribed - 50:1 in the case of BT here in the UK.
Regardless of any QoS scheme the carriers put in place, there isn't enough bandwidth for video to operate in these networks. VoIP maybe, but not video - even if it is crammed into a 1Mbit/s Media 9 stream!
Cheers,
Geoff