| I attended Cisco's (NAG, Architecture
Geeks' Conference) a three day meeting at Cisco of lead Cisco
engineers, engineers from the big networks, and a few researchers.
There were more Treo 650s in the room than coffee cups.
The Cisco IPTV path was shown in the most complex way possible.
It appeared as if one had to run MPLS, DiffServ, and every other
gadget Cisco has ever considered running IPTV over your network.
Though this is actually not the case, one left not knowing how
to deploy IPTV simply and incrementally. The object seems to
be selling complexity to sell more hardware.
- Pankaj Patel, VP Router Division, tried to probe the 'convergence'
space, though seemed to have more than three kinds of convergence
in mind. He did take a break from complexity to note that scaling
and speed were of interest. Ted Seely, Sprint, asked when the
CRS routers would actually work, have enough line card power,
have enough RAM, etc. These have been issues on all platforms
for over five years. Patel dodged.
- Wen Chen, Cisco Fellow, gave us a lengthy tutorial on video
encoding standards, algorithms, and techniques followed by his
new very clever compression and encoding algorithm. This appeared
to be solid research.
- Nandita Dukkipati of Stanford described RCP, their team's
research into congestion management protocols beyond TCP. Aside
from being a much simpler algorithm than a well-known competitive
algorithm, XCP, it performs a lot better. On the other hand,
it is only useful when it is used at the most congested link
in the path, as it is not based on dynamic sensitivity to flows.
See http://yuba.stanford.edu/rcp/.
- Yashar Ganjali of Stanford presented his group's work on
radically reducing buffer sizes in routers. His NANOG presentation
pretty much sums it up http://www.nanog.org/mtg-0510/ganjali.html.
- Jennifer Rexford of Princeton described the current NextGen
Internet research initiatives in NSF and DARPA being driven
by folk such as Dave Clark (MIT) etc. The assumption was that
the current internet is too rigid to be able to experiment and
develop an next generation data network. This assumption is
looked at with some skepticism by folk in the operational internet,
as this was the theory behind Internet2, Abiline, .. See http://www.geni.net.
- Dave Meyer rambled on in a complex fashion about complexity.
- Vineet Mehata described the US government efforts in military
architecture and migration using the Transformational Satellite
Communications System and the Global Information Grid architectures
as examples.
- Maria Napierala of AT&T told us of their used of multicast
over MPLS 2547 VPNs. A wonderful exercise in over-complexity.
- Stafano Previdi of Cisco described IP Fast Reroute as implemented
in the IOX/XR platform. It uses no signaling, aka protocol changes,
just using what is already in the link database. There are a
number of proposals, but this one was Loop Free Alternates,
having a backup destination for each next-hoop link (not prefix)
in the database. There is no need for flag days. It can be partially
deployed. It is way cool! |
