Comments on: Broadband dreams and multicast 'beams'

Internet pioneer Vint Cerf says the Internet will only make good on its true broadband promise when ISPs finally adopt multicast technology.

[CAOgdin75]

"20 kb/s" Is Bogus

Vint Cert claims "...only an average of approximately 20 kilobits per second for each subscriber." I'm curious where he gets this outrageous claim.

On my 384Kb/s DSL, I consistently get over 300 kb/s throughput. On my nominally 1.5 MB ADSL, I consistently get over 1 megabit.

20 kb/s is less than half the speed of 56K modem.

I just don't believe it, 'cause if it was true, there'd be big class-action "bait-and-switch" litigation.

You are missing the point

Vint Cerf's point was that consumer ISP's over-subscribe upstream bandwidth on a ratio of 1/50.

I agree w/ his opinion that multicast needs to be implemented across consumer internetworks. The current network topologies and business models prohibit content owners from providing higher quality content. The flipside to this argument however is that ISP "x" would potentially have to either perform a massive infrastructure upgrade to support multicast or even forklift existing infrastructure.

Mr. Cerf you would be surprised to know that there are upstream providers and content owners/distributors that are ready for downstream providers to support multicast.

Hold on a second.....

He said, "they make available only an average of approximately 20 kilobits per second for each subscriber." Here is my take on what he means.

If you were to get all of your neighbors to actively downloads or browse at the same time, the would have to split the bandwidth up depending on how much bandwidth is reaching your neighborhood. On an average, his claim is very true. Although you say you do reach the bandwidth that your promises you, there is no way you would reach that if you were in an area where many people were using the at one time. An example would be New York. The bandwidth in a city as active as this could very easily be somewhere around 20 - 40 kilobits per , because this is all your has dedicated to you. With all of this said, his claim is correct based on his discussion of movie . If the replaced the local movie rental store, then there would be millions of people downloading movies. If this were happening at this exact , the would be very slow. are not ready for this kind of change. Yes, you may your bandwidth and then some more, but if the load is intense, your has really only dedicated you 20 kilobits per second (download speed, not throughput or downstream).

Also, about the bottlenecking of the speed. "" can be split, but this would not help if the was not ready to split the . The speeds today are just not fast enough, and would by far surpass your opinion sir.

You can't do Math - Do you understand Multicast

The article does not say 20kbs a sec = it tries to explain multicast - though it misses the issue of VOD considerations.

Multicast vs. on demand

In order to be truly an 'on demand' service, every video stream has to be controlled independently,
and there is no 'downloading' to a local hard
drive before the watching starts. This fact
completely rules out multicast as a solution,
because every video stream needs to be controlled
individually.

In Asian countries (like Japan) subscribers can
get symmetric 20 Mb link for a bit less than US
customers pay for their 'big' 3 Mb service. The
concept of broadband is very relative. Anyways,
with a 20 Mb link, there is no issue in getting
HD-TV service, in addition to phone and internet.
The technology for such does exist and we must
not think that the US has is the best example on
what can be done with it. At least, not yet.

Well...

20 mb/s would not help us in this situation. You contradicted yourself by saying that we would need an "on demand" service which would mean not downloading to a local hard drive. How would 20 mb/s help this? The streaming would not be fast enough for DVD quality movies. Also, you must completely understand multicast in order to understand the concept of integrating it into the internet.

Broadband dreams and multicast 'beams'

Multicast broadband is a huge success in Italy. Fastweb (www.fastweb.it) offers tv-broadcasting and video on demand services using standard ADSL technology

Hold on a second....

Hey said, "they make available only an average of approximately 20 kilobits per second for each subscriber." Here is my take on what he means.

If you were to get all of your neighbors to actively downloads or browse at the same time, the ISP would have to split the bandwidth up depending on how much bandwidth is reaching your neighborhood. On an average, his claim is very true. Although you say you do reach the bandwidth that your ISP promises you, there is no way you would reach that if you were in an area where many people were using the internet at one time. An example would be New York. The bandwidth in a city as active as this could very easily be somewhere around 20 - 40 kilobits per seccond, because this is all your ISP truely has dedicated to you. With all of this said, his claim is correct based on his discussion of movie transfering. If the internet replaced the local movie rental store, then there would be millions of people downloading movies. If this were happening at this exact momment, the internet would be very slow. ISPs are not ready for this kind of change. Yes, you may recieve your alloted bandwidth and then some more, but if the load is intense, your ISP has really only dedicated you 20 kilobits per second (download speed, not throughput or downstream).

Current version of Multicast is useless for most of applications

Multicast is a nice thing in theory, but in practice it has severe drawbacks. Multicast, which is configured at today's routers, has no any broadcast group control features. Anyone can connect to a broadcast and i.e. download a movie. Moreover anyone can send to an existing Multicast group. Just imagine what will happen if someone, even by mistake, will send a big data stream to a large broadcast group. It will flood links of thousands of users. Another problem of Multicast is QoS. The Multicast uses unreliable UDP protocol, which is becoming even more important issue with growing popularity of WiFi networks. To deliver high quality movie all packets have to reach destination. This requires packets repetition for particular clients. So Multicast doesn't scale so nicely as advertised. Of course above are just some of Any Source Multicast drawbacks.

[msansing]

Multicast won't do it

In Mr. Cerf's model viewers will be limited to what can be cached in their set-top box. If we're talking high-def video the storage requirements are quite large and therefore the amount of video they will be able to watch will be limited--probably something like the 10 or 20 movies offered on PPV services today, and nothing like the promise of unlimited selection that the internet could provide if bandwidth issues are addressed. You could multicast a tremendous amount of content and let users just cache what they appeal, but I don't know how feasible that is.

Of course I think there is a roll for multicasting--today's broadcast and cable networks could migrate to the internet and multicast there...there will always be a certain amount of content people will want to watch at the same time e.g. live events, breaking news or stuff people will want to discuss over the veritable watercooler the next day. But I think you need both the on-demand option with a wide selection plus a limited feed of multicast programming.

Multicast Does It

What you suggest about ?multicast[ing] a tremendous amount of content and let[ing] users just cache what ? appeal[s to them] is completely feasible. In fact, we are doing this right now at Wavexpress.

Over any network that is IP Multicast-enabled, Wavexpress Multicast channels offer different genres of content. Users subscribe to whatever channels interests them. Programs for those channels are cached and later deleted according to rules set by the content publisher. Users also have complete discretion as to how much content is cached and, of course, greater granularity is possible such that a consumer could elect to only cache a portion of the content on a particular channel.

One of the great advantages to cached IP Multicast, besides the greatest bandwidth efficiency, is that you can surf the web with no degradation. For example if you are a cable modem user and your cable operator has provided you a 1.5 Mb/s connection, the Multicast packets do not count against the throttle. So these packets do not slow down your regular (unicast) web surfing and the Multicast content is cached transparently. Additionally, this service works great across wireless LANs because the content is packaged in files which employ forward error correction so you can have substantial packet loss with no loss in video quality.

This needs clarification!

Dearest Vint,

1) The 20Kbit shared access is only from the ISP to the global
internet, and it requires all 50 to be online at the same time -
which they of course never are. With a 10% concurrency, the rate
is suddenly a respectable 200K.

2) But video and TV programs are likely to be streamed from
servers within an ISP network, and thus only depend on the line
to each household. DSL can in most cases do more than 2mbit
but even up to 8mibt, and the new ADSL2+ can do up to 20mbit
or more. Cable modems can do 40 mbit/channel and with 80
channels the capacity is 3.2 GB for each loop. And loops can be
made smaller by inserting new headends in the tree network.

3) Compression means less bandwidth is required to transmit
DVD-quality video and TV. With the new H.264, 1-1.2 mbit/s is
considered DVD quality.

4) Multicasting not useful for video as it requires local storage
with is impractical and not desired by content providers. But
multicasting is key for live TV transmissions, which can give
cable, satellite and terrestrial broadcasters a serious run for
their money.

5) There will be multiple broadband paths into the home. WiMax
ISPs will start operations next year and offer DSL-like
performance at perhaps even lower prices and different
coverage. Hence, each network will not be maxed out at all
times.

boardcast beams and multicast dreams

Vint Cerf,
Is this something that could be done? tested? anytime soon at a
Public Access TV stations, which have limited distribution only to
a local area via cable? Could their programs be made available
to a wider audience this way?
Thanks,
Hal

[soul_tech]

Well.. Not really

In DSL and Cable modems, there is a 'shared' link from the ISP to your neigborhood, called the "aggregation" link. This link creates 2 problems:

1. It's not typically IP routed, but ATM based, so it doesn't understand IP-based (typical) mulitcast.

2. It is oversubscribed typically at 4:1 minimum, but often much higher than that, meaning that if you have a 1.5Mb/s link, then 4 or more neighbors will share an aggregation of 1.5Mb/s. Ratios of 50:1 or greater are not uncommon.

The idea has promise, but still has the typical "last-mile" problem.

20kbs - is the Writer a former Exec at a compresson scam company ?

Shannons Law everybody - Video to the computer - COMPUTE THE REQUIRED BANDWIDTH TO SUPPORT SUCH then calculate the infrastructure costs. Again - the wrtier should stick to numbers not technology. This is NOT NEW TECHNOLOGY - MultiCast is NOT A TECHNOLOGY ISSUE - and it will NEVER get past HOLLYWOOD until all the DRM issues are resolved. TV and Movies - THE NOISE ONLY GETS LOUDER. I HAVE CABLE - I HAVE A CABLE BOX WITH A DVR FEATURE - WHAT DO I NEED THIS FOR AGAIN - ADULT ?

[rkhalloran]

Satellite feed? Already tried..

With the dot-com bust, the companies that had tried satellite feed, most notably Skycache/Cidera, folded in the face of telco's dropping prices to try and sell all that landline bandwidth they'd strung up during the boom.

Now with the broadband boom there *may* be enough demand to revisit this, but I expect any potential investors would be gun-shy at this point.