Telepocalypse

Martin,

Towards the end of your post, you suggest that once people have a thick pipe running into their homes, they will download video directly from non-telco video servers, cutting into the 'triple-play' revenues that are the primary driver for the telcos to create the broadband infrastructure.

I don't think that circumventing telcos' survival strategy will be so easy as pointing your browser to the url of an independent video server.

Based on what I hear from product managers at equipment vendors participating in the IPTV RFPs, extending QoS-aware IP networks all the way to the end application is an integral part of the triple-play strategy. Without QoS-aware IP networks that are end-to-end -- and not just limited to the core -- it is impossible to deliver Video-over-IP that comes anywhere close to the analog TV we all are used to.

Now, while implementing these QoS mechanisms in the core, edge and home network, the telcos want to make sure they provide higher, video-like QoS only to applications they can trust. In other words, only a set-top box that connects to the telco's video server will get the 20 mbps bandwidth; traffic to other sites, including independent video servers, will be relegated to the best-effort category, making it unsuitable to stream videos from non-telco video servers. At least that's what the service providers are gunning for.

Apparently, Cisco has been able to 'tag' Skype packets, so you can rest assured that any telco with a VoIP offering that deploys Cisco gear will configure its routers to 'randomly' drop just enough packets in each Skype call so that its own VoIP offering has a better audio quality than Skype. P2P traffic is another target of the telcos' traffic shaping (read: bandwidth discrimination) efforts.

In short, the telcos and equipment vendors are working hard to ensure that not only can they distinguish between trusted (read: revenue accrues to service provider) and non-trusted user applications, but also control bandwidth distribution among these applications.

If the telcos get this right, it becomes clear that all non-telco VoIP offerings are short-lived. In fact, as gatekeepers of the thick pipe that delivers triple-play services to a home, the telcos may perhaps end up being more powerful than as providers of PSTN telephony. So much for VoIP representing the demise of the telco!

Note: Mechanisms to implement QoS and traffic shaping will necessarily reside in the modems, routers, hubs, set-top boxes that will make up the home network. Obviously, a tech-savvy consumer could build his own network from scratch, thereby making it possible for even rogue applications to grab Video-QoS from the network. In this case, it would certainly be possible to stream video from non-telco video servers.

But most consumers, who do not want to debug a home network, will go with the equipment provided by the telco, especially when it collapses into a single box (2Wire packs a router, firewall, hub, wireless AP and ADSL modem into a single unit), in which case the telcos will indeed end up as gatekeepers of the 20 mbps broadband connection.

Posted by: at November 11, 2004 10:19 AM

Ash, I suggest you go read David Isen's original essay about Dumb networks. With fiber, there is no practical limit to the bandwidth available to a home. Given available standards, every home could have a GigE link, which is what many new corporate lans have as an option today. This pricing Martin refers to above reflects the telco's effort to exploit the current market perception re: bandwidth scarcity while also extending that perception. Once the telco go FTTH, the cable cos will follow suit and price per bit will continue to plummet. It is an inescapable consequence of the technology. With all-optical networks, there is no need for intelligent features in the network's core - they just degrade performnance and increase cost. All i want from either provider (cable or telco) is a very simple and highly commoditized product - an ethernet connection.

Regarding the IPTV vendors, I am guessing those platforms must be architected for the lowest common denominator within the "broadband" network, which would be some form of DSL (which would also be the most typical form of transport available for quite a while)

Posted by: at November 12, 2004 12:00 AM

I don't follow the logic of overcharging for 30Mbps access. Can we not "inverse multiplex" 2 15Mbps links using a simple load balancing algorithm?

Posted by: at November 12, 2004 05:33 AM

Any FTTH service is going to come with hard bandwidth caps that will prevent TV over the Internet from being useful for most people.

Bandwidth within the FTTH network is pretty cheap, although not unlimited. The FTTH provider can shape the video network traffic laod using multicast to keep the max data rates manageable.

Internet bandwidth is both limited and (relatively) expensive.

The FTTH provider can't provide unlimited internet bytes without charging more for it.

Don't forget that 20 mbps at the edge doesn't mean that the network can ssutain that rate - network bandwidth is way oversubscribed because most of use are using low bandwidth applications like reafig email and surfing the web. If everyone was trying to use the internet for watching TV, the network can't sustain the internet traffic load.

Watching IPTV originating within the network is easier bacause IP multicast will reduce network load. Note that single stream/VOD applications will oftetn still be netowkr limited in how many stream can be handled at once.

Posted by: at November 12, 2004 05:03 PM

I expect the next step to be DVRs (and P2P networks) that are network topology aware. You only need to burst a file through the Internet uplink once into the local network and you're done. People will then cease to ask "what's on TV" and instead wonder "what's available locally". And with cheap multi-terabyte storage happening within the amortisation period of these networks, that's a headache for the network operators.

I also suspect that multicast is a doomed technology in the long run (>10 years) as any scarcity of bandwidth is an economics problem, not a network problem. You solve "layer 8" (finance) problems with layer 7 solutions. Your DVRs will have to collaborate to ensure you all get to see the superbowl. But you won't know or care.

Of course, DRM pisses in the flowerpot of connectivity, but that's another story.

Martin

Posted by: at November 12, 2004 05:11 PM

Just to elaborate the last point, which is more attractive? A choice of a hundred channels, or a choice of a million shows?

Some of these technologies don't in fact need to threaten the telco. Once you renounce the religion of control, you suddenly see that there's a role for telcos to coordinate these ad-hoc networks, and indeed negoatiate the DRM maze and arrange payments for what flows through the P2P networks. This, naturally, requires a degree of enlightenment not commonly found in telco boardrooms today.

Another wildcard here is the possibility of competing fiber. In the exurbs of America you might not get to see it because of the low population density and incumbent political power. But in densely populated areas elsewhere you can pass a lot of homes by digging down one street. Then things get interesting. Remember, not everywhere is as backwards as the US ;-)

Posted by: at November 12, 2004 05:39 PM

Pardon me sirs, but I don't understand the math about the PC remote backup. Having a 250 Gbytes drive doesn't imply 250 Gb to backup daily. For example, the departmental server in our consulting company has a capacity of 40 Gbytes, but *incremental* backups weigh around 200 Mbytes. This corresponds to the data we produce daily (new versions of documents we create or receive, outgoing and ingoing emails...). Even 24h of TV viewing produces at best 20 Gbytes. (Forget about HDTV: people are satisfied with plain old TV ; they will use HDTV if it's the *same price* and *not less convienient* -- look at CD, then SuperAudio CD and MP3 for proof).
So don't overestimate the need for speed on file exchanges which are the base of every Internet protocols: smtp, ftp, http, p2p... And for broadcast and simulcast of video channels, didn't we already have some networks intrinsically designed for that (satellite, cable), hence cheaper?

Posted by: at November 17, 2004 08:56 AM

I hope you're testing your disaster recovery often, because chained incremental backups are a uniquely wonderful way of increasing the recovery failure rate. Personally I'd expect to do a full backup on at least a weekly basis. If you assume daily incremental backups, and monthly full backups, and a .1% failure rate for any one backup, then the cumulative chaining of 31 backups gives you a 3% chance of non-recovery. Unacceptable in my mind. Incremental backups are also more liable to operator cock-up.

As Andrew Odlyzko observed, file transfers indeed dominate streaming, and increasingly so. But that message hasn't sunk into the telcos, who are stuck trying to turn streaming into their next cash cow (e.g. Sprint's mobile TV service). So the pricing of fiber may be driven by a business model mirage.

Interestingly, the broadcast flag could indeed poison the market for HDTV, since, as you note, it makes TV less convenient.

Posted by: at November 17, 2004 01:18 PM

Indeed we make full backup weekly, etc. But I was not speaking of differential backup with some clever tools that break files into pieces and copy only the deltas. We are using plain duplication of those file which were 'touched' during the day. It works very well with programs that expose their granularity to the filesystem (don't start me on Outlook .PST against maildir IMAP storage).
Added benefit as we are storing at least 15 days : we can restore "the version of that document we were working on Monday"... that was trashed totally (the maniacal assistant) or partially (the cut-and-never-paste boss).
On the long run, this arrangement rests on the assumption that storage is cheaper/has more capacity than long distance telecom. My reading of Christensen convinced me of that.

Posted by: at November 17, 2004 03:45 PM

Indeed we make full backup weekly, etc. But I was not speaking of differential backup with some clever tools that break files into pieces and copy only the deltas (although rsync is a mature solution). We are using plain duplication of those file which were 'touched' during the day. It works very well with programs that expose their granularity to the filesystem (don't start me on Outlook .PST against maildir IMAP storage).
Added benefit as we are storing at least 15 days : we can restore "the version of that document we were working on Monday"... that was trashed totally (the maniacal assistant) or partially (the cut-and-never-paste boss).

On the long run, this arrangement rests on the assumption that storage is cheaper/has more capacity than long distance telecom. My reading of Clayton Christensen ("The innovator's dilemma") convinced me of that.

Posted by: at November 21, 2004 07:13 PM