Re: [EFM] Network timing, ATM, ADSL/VDSL and EFM
On Fri, 28 Sep 2001 10:47:47 -0700 Zi Wang wrote:
> The problem with EFM is that it is difficult to provide voice and video
> service using EFM. ATM, on the contrary, has proven track record on this.
> The small cell size of ATM, makes it possible to translate voice and video
> in a timely fashion. The bandwidth tax on ATM is insignificant if you can
> get 138MPBS from 155MPBS, who cares the 16MPBS tax. You are billionaire and
> you do not bother to pay some tax.
The figures I trust are for IP performance. The IP engineer rule of
thumb is IP over ATM is roughly 70% efficient. That is, if you pay
for 155mb/sec, you should expect to get about 110mb/sec. Also, the
word on the street is the failure mode for IP over ATM is real nasty.
If you ever max an IP over ATM link out, it will crash and burn
instead of degrading gracefully.
I don't expect you to care about or believe my opinion expressed
above. By the same token, understand that this opinion expresses 10
years of pain and suffering on the part of many ISP engineers.
I have zero interest in hearing a explanation about how our
experiences are not valid. If you get in a bar full of experienced IP
engineers, I would recommend not bringing up the subject of ATM.
So we use ATM any time its price drops below 70% for a compariable
fixed bandwidth circuit. ATM is not evil; it is just not efficient
> If EFM is only for data, it is a better choice than ATM. But if EFM is to
> provide data, voice, and video, I doubt. Any people have idea what the cost
> to provide video, voice, and data by EFM compared to ATM ?
I could not agree more! This was the point I was trying to make. If
you want timed service to provide data, voice and video over the same
circuit, the existing ATM protocols are the best protocols I know of
to provide those services. I don't understand why people think we can
take the same design goals as ATM for EFM and do a better job this