Re: Real Media and M-Bone feeds
Fredrik Nyman <fredrik@swip.net> wrote:
Multicasting cannot be made scalable. It is as simple as that.
Think posivite! And see comment below.
I love that. Do you seriously think that positive thinking can change the fundamental problem of having traffic-dependent global routing information? We're having lots of problems with traffic-independent routing information already, thank you.
How are you going to handle 100000 realvideo streams? If all those 100000 sources are in the US and all recievers are in Europe, there would be no difference in the amount of bandwidth required across the ocean with multicast compared to realvideo.
Easily. It is called cacheing. Multicasting provides exactly the same bandwidth savings as cacheing with cache retention time set to zero. Now - how are you going to handle the same users who want to watch newsreel from the beginning, so as not to miss the important news they caught the end of? We already have cable companies transmiting the same content with different delays many times over. That's one of a hell of efficient technology.
And I'd definitely prefer to use multicast when I bring all the worlds radio and TV-channels to all the worlds users.
Yep. Sure. The only problem that _nobody_ needs or wants simultaneous transmissions. Just because it is the common practice due to medium limitations, it does not mean it is desireable or has any market potential. We ain't need no second television. The old one is cheap and widely deployed, thank you. Note that i left the issue of oligocasting (multicasting to few parties, such as useful in conference calls) out of the picture. Oligocasting at L3 has even worse scalability problems, and the only realistic approach at this time seems to be using MCUs - like the telephony people do for decades.
Given time and enough resources multicast will rock.
I would prefer backbones to work steadily, without any rocking motions.
Would you have thought 10 years ago that there'd ever be something called world wide web, and a zillion users on the Internet?
10 years ago i definitely was a believer - starting the first ISP in a country spanning 11 time zones :)
How could you possibly scale the Internet using IGSs and the occational AGS?
Can't. 12000s and M-40s aren't much better, either. That's why i invented the terabit routers :) --vadim
Thus spake Vadim Antonov
Fredrik Nyman <fredrik@swip.net> wrote:
How are you going to handle 100000 realvideo streams? If all those 100000 sources are in the US and all recievers are in Europe, there would be no difference in the amount of bandwidth required across the ocean with multicast compared to realvideo.
Easily. It is called cacheing. Multicasting provides exactly the same bandwidth savings as cacheing with cache retention time set to zero.
With the constraint that if you want to provide exactly the same bandwidth savings, you have to have a cache at every hop. Without a cache at every hop, there still can be significant savings, but not quite equal to the bandwidth savings of multicast at every hop. (Though, admittedly, multicast capable routers aren't available at every hop either). -- Jeff McAdams Email: jeffm@iglou.com Head Network Administrator Voice: (502) 966-3848 IgLou Internet Services (800) 436-4456
Easily. It is called cacheing. Multicasting provides exactly the same bandwidth savings as cacheing with cache retention time set to zero.
With the constraint that if you want to provide exactly the same bandwidth savings, you have to have a cache at every hop. Without a cache at every hop, there still can be significant savings, but not quite equal to the bandwidth savings of multicast at every hop. (Though, admittedly, multicast capable routers aren't available at every hop either).
Except that you probably don't care about every hop either (who cares if the 100Mbps to my access server with 4 T1's on it is partially wasted?). -Phil
participants (3)
-
Jeff Mcadams -
Phillip Vandry -
Vadim Antonov