> From what I see, there seems to be a lot of emphasis on GigE and
>the very rapidly upcoming 10GigE combined with QOS now adays. ATM
>really doesn't seem to work that well with data (TCP/IP) and has a very
>high overhead. Ethernet is getting fast enough that when combined with
>QOS it can easily handle voice and video as well as data. Also,
>ethernet is cheap, cheap, cheap; even GigE when you compare it with
>ATM, and just about everybody knows how to handle ethernet, but ATM is
>something that relatively few people know really well.
>
In the IETF and similar forums, you'll often hear MPLS called "ATM
without cells." As transmission rates become faster, some of the
problems that ATM was intended to solve go away. When an OC-192 (10
Gbps) medium can send an entire frame faster than an OC-3 might send
a cell or a few cells, the better interleaving of cells becomes a lot
less important.
Also, forwarding components have been getting faster, and the fixed
cell length is less important in achieving a fast forwarding rate.
You still need to look at the IP header to set up MPLS label switched
paths (LSP) and to assign a packet to a LSP at the edge, but label
lookup and forwarding has comparable performance in ATM and MPLS.
Photonic switching, where traffic is rerouted based at the high-speed
stream level rather than the packet or cell level, isn't here in
production, but it is coming rapidly. Photonic switching will
complement, not replace, routing. Please do not get me started on
the buzzword of "optical routing." With the capacity of newer
optical transmission, bandwidth becomes much cheaper than forwarding
logic, and some conventional bandwidth-conserving methods simply
won't be important in large networks.
To give some perspective, OC-192 (approximately 10 Gbps) is getting
to be reasonably standard in carrier networks. Juniper routers have
had some lead on Cisco in doing efficient forwarding at these speeds.
The next speed level is OC-768 (approximately 40 Gbps). Chip
designers don't have immediate strategies for increasing single
stream speed above this rate.
On the other hand, the Dense Wavelength Division Multiplexing (DWDM)
people can carry multiple OC-192's and OC-768's on a single fiber,
and switch these streams today with partially photonic components.
Transmission systems that can carry 160 OC-192's on a single fiber
are commercially available, and my transmission colleagues tell me
that 320 streams will be available in the near term. DWDM will be
available for multiple OC-768's on the same fiber, and, of course,
carriers install more than one fiber at a time.
For those of you that subscribe to IEEE Spectrum magazine, the
current issue has some nice overview of these and other emerging
technologies.
_________________________________
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
