Hey all,
A few comments on the draft:
============================
When cataloging the advantages of satellites over fiber, you might want to
mention satellites' ability to support mobile users. A ubiquitous
satellite network will enable internet everywhere not only on the
ISP-scale, but also on the personal scale allowing anyone, anywhere,
anytime network access.
For these mobile satellite users, it might be possible to take advantage of
the extra information that: 1) yes, there is definitely at least one
satllite link involved and 2) we know that the first hop is a satellite
link. Significant performance increases could come through coordination
between the mobile user and the earthstation he's downlinked through.
These methods would probably fall into the "further research" category.
Has any thought been given to ranking the desire for the various
standards-track mods in different user environments? For example, mobile
users (at least for now, sigh...) don't need big windows because data rates
are too low. SACK and Fast Retransmit&Recovery, on the other hand, are
life-savers due to the (even higher) BER of the mobile satellite
environment.
JPL's Mobile Satellite Testbed:
==============================
Also, at JPL we have our mobile satellite protocol testbed up and running.
The testbed uses propagation data taken from ACTS and TDRS to drive a
channel emulator. One end of the simulated satellite link is connected to
a LAN at JPL with internet access, while the other end is used to simulate
the mobile user.
With the testbed we can examine the performance of all combinations of the
mitigations mentioned in the draft, plus some others, at data rates up to 2
Mbps and GEO delays. Thus the performance of, for example, SACK with Fast
Retransmit & Recovery, can be looked at as a function of the channel BER,
which is influenced by the amount of FEC overhead.
It looks like the UCLA cs group has looked pretty extensively at pure SACK
implementations, but I hope to post some results on window scaling, FEC,
and combination experiments soon. I'm particularly interested in TCP's
performance at moderate to high BER (and its improvements with varying
amounts of FEC and SACK) and in implementing some of the explicit
corruption notification schemes recently mentioned.
If anyone has ideas for other tests which might be more directly applicable
to the draft, I'd like to talk to them about it.
--keith
-----------------------------------------------------------------------------
Keith Scott [email protected]
Jet Propulsion Laboratory
4800 Oak Grove MS 161-260 (Voice) +1.818.354.9250
Pasadena, CA 91109-8099 (FAX) +1.818.393.4643
-----------------------------------------------------------------------------
This archive was generated by hypermail 2b29 : Mon Feb 14 2000 - 16:14:31 EST