> > If response is faster with IPTV, it must mean that > > the IP multicasts are flowing throughout the IPTV > > network all the time. Because you can easily > > expect similar delays for IP multicast, if you > > expect to be building and pruning multicast trees > > as users "tune" in and out. Therefore, in practice, > > the bandwidth is not really being managed all that > > differently. The core of the IPTV net will be > > loaded with TV streams, much as cable nets are with > > their fixed 6 MHz bands. Craig Birkmaier wrote: > > VOD is not "typically an IP multicast service. Yes, true, unless it's NVOD. I was really describing the situation with live multicast streams, when IPTV subscribers flip between channels. > You can route everything from the central > office/head end, in which case the network will > only contain packets being routed to specific STBs. > Or, you can place routers close to the home and > deliver the most popular multicasts out to the edge, > preserving "trunk" bandwidth. The content needs to get to those servers close to the edge, which takes up trunk bandwidth. But certainly, for VOD, that's the way to go. Or hey, better yet, why not locate the storage on the customer premises? That's the tradeoff. With cable, while the network can in principle install a distributed set of servers for better VOD (and migrate that to IP), this isn't mandatory. The PVR function can be owned or rented by subscribers. With IPTV, instead, it's more appropriate that the network incorporate those servers, because the drop cable to each subscriber is not very wide. To me, that's not a slam dunk advantage of IPTV. It's more like a work-around. However, *if* service providers prefer to incorporate their own PVRs within the network, because it provides them with more revenue, then you might conclude that the IPTV last drop bandwidth constraint is not a problem. To the provider, might not be. But to the consumer? > with MP{EG-2 TS based system you can waste > considerable bandwidth padding out the substreams > within a 6 MHz channel (although stat muxing of the > streams helps to maximize the utilization of each 6 > MHz channel. Remember that with 256-QAM, you get 38.8 Mb/s out of each 6 MHz channel. That gives you plenty of bandwidth for doing efficient stat mux, much as an Ethernet link would do. And of course, DOCSIS puts IP traffic over these 6 MHz channels, so you can do anything permitted by RTP/RTCP just as well over MPEG-2 TS as you can over Ethernet. Including resynchronizing multiple incoming streams into one output stream. The differences are just not so big. Not if cable and IPTV systems are each configured as well as they can be. > They are very different. IP multicasts are a SOURCE > for the routers in an IPTV system. The viewer > chooses to join a multicast, which is then routed to > that home. How far the multicast is carried through > the system infrastructure is just a question of > system design. Yes, that last sentence glosses over what I've been talking about. Imagine a market full of couch potatoes flipping between live channels. Each subscriber conducts IGMP with the edge router of the IPTV net. If you go for bandwidth efficiency, the edge router would then conduct a multicast routing protocol to request or drop membership to these multicast groups from the network core. But these protocols don't react instantaneously. They can't. So typically, if group membership is dropped by all IGMP hosts in a given subnet, the edge router will continue to receive the multicast packets of that multicast group from the core, for a period of minutes. Until that tree branch is pruned. This means that as subscribers flip channels, the core is becoming saturated with multicast streams. They aren't being dropped instantaneously, even if no one is watching *right now*. Also, if you take the max efficiency strategy, a subscriber joining a multicast first within a given subnet will have to wait for the network to graft this new branch onto an existing multicast tree, or might even have to reach as far back as the source edge router to start a new tree. This can easily take 20 or 30 seconds, depending on the size of the network and on how much background traffic (overhead) the network admin is allowing for multicast routing protocols. The faster the response, the more the overhead. So in practice, because of the two effects described above, IPTV systems don't shoot for super bandwidth efficiency. They compromise for a better viewer experience. The live multicast streams take up core bandwidth, so there is no real possibility for infinite channel selection, or anything remnotely resembling that. For VOD, you have the problem of network saturation, just as any unicast streaming media network has. Location of servers close to the edge helps, but it adds cost too. And again, servers have limits, so there is no possibility of infinite channel selection, or anything close to that. So in practice, no big difference. Bert ---------------------------------------------------------------------- You can UNSUBSCRIBE from the OpenDTV list in two ways: - Using the UNSUBSCRIBE command in your user configuration settings at FreeLists.org - By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word unsubscribe in the subject line.