Stephen:
Thanks for the links.
Here's a fairly recent overview of 10Gbe. Its not really what you're looking for,
but may be of interest nonetheless:
ftp://ftp.netlab.ohio-state.edu/pub/jain/courses/cis788-99/10gbe/index.html
This is a fairly good review, and by a .edu. It is dated November 1999 and is a bit
out of date, but still worth reading. You are right, they don't discuss traffic,
particularly not multi-access multi-media multi-hop mutli-server/client architectures.
An assortment of white papers on the Last Mile for Broadband. One by Cisco
on ethernet over fiber. Disappointing. They basically proclaimed that this is what
would be:
www4.nationalacademies.org.
You are right, the CISCO paper tends to be a proclamation. However, the first
paper on the site" by George Abe, Palomar Ventures , "Factors Influencing
Investment in Residential Broadband Equipment and Services - a Venture Capital
Perspective", is quite good, and gives a good summary of the venture view of the
domain- which is also similar to the stock investor's view.
Here's a free seminar put on by Network World and Yipes in 10 cities around
the country:
networkworld.com
Here's another site I just recently ran across. The number of technical papers here
are massive. You might want to nose around here for awhile:
tradespeak.com;
It is a site giving whitepapers in the 'battle of the forums": dsl vs. atm vs. gigabit
ethernet. A lot of good stuff, a lot of slanted stuff, and a brief introduction to
Multi-Protocol Over ATM (MPOA)
atmforum.com.
Which leads me to my next point.
As noted, I''ve been looking at how gigabit ethernet, or any collision detection
scheme, or any network protocol can handle multi-everything traffic. The best
approach seems to be to use Multiprotocol Label Switching (MPLS), both for
ATM and IP. MPLS is a labeling scheme, not a priority scheme, and allows you
to, among other things, make virtual managed IP networks on a hop by hop basis,
a sort of SVC for IP. Sounds good to me.
ietf.org
mplsrc.com
have a lot of information on MPLS, and how it helps solve some of the
multi-protocol network traffic problems, even if you are sort of duct taping an
ATM protocol under IP packets.
In addition, this thread has a number of recent posts (using the keyword search),
including Frank and Gary's dialogues:
Message 14328841
Message 14329261
Basically, you can now create separate MANAGED IP metro/core networks after
the first hop (the head end for cable, the CO/DSLAM for DSL, for example), by
leveraging the simple MPLS signaling to diverge traffic.
What I like is that each separate managed network can now be engineered to
handle the traffic independently, depending on demand type and traffic load: so the
metro and core will be a set of independent single media networks that grow
independently. You can also use different network protocols via gateways to haul
traffic from that point on: say TDM for voice, ATM for video, IP for internet-
whatever works for you, the network provider.
The really nice thing about this is you don't have to place you bets on technology
and capacity up front- you go with the flow and build your network when you need
to.
Two issues with MPLS:
-the multi-media, mult-host/client traffic problems remain in the last mile
- I'm wondering about the overheads for some types of IP traffic (overheads are
usually the problem in self routing control networks)
An VOIP packet contains a IPv6 header (60 bytes- v4 is 40 bytes), a UDP
header (8 bytes) , a RTP header (12 bytes) and a H.729a payload of 10 or so
bytes, and MPLS (4 bytes). This is 84 bytes for a 10 byte payload. Now even
H.729a is not quite as good quality as 'toll' quality voice, and it is about, I
remember a 10:1 compression algorithm We are almost at parity with TDM! in
bandwidth efficiency and getting lower quality.
Now what if we had fiber to the desktop, or multi-media set top, and ran ATM
everywhere- doesn't this solve all the worlds problems? The other approach is to
continue to solve multi-media access problem by duct taping some more ATM
management protocols onto IP (they are working on this, apparently), but then we
are getting, well, ATM with varying length cells.
Perhaps it is time to go back to that idea of ATM to the desktop (or set top), and
LAN/IP/ethernet in the home. I suppose this is too radical to overcome the voip
momentum, and we do need multi-line and voice over cable capability that only
voip can provide in the DOCSIS standards. But it may indicate that ATM
protocols will stay in the metro/core for a long time, even if we call it IP (none dare
call it ATM!).
I guess from previous posts the thread has mostly decided can keep our SONET
chips stocks for a while; it looks to me like we can keep ATM chip stocks as well.
justone opinion |
