[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [RRG] Are we solving the wrong problem?
Hi Jari,
El 04/03/2008, a las 12:13, Jari Arkko escribió:
Mark,
These are interesting thoughts and worthwhile to be looked at in more
detail. Do you have a proposal?
You are basically going into the direction that is similar to Shim6 in
the sense that multihoming is exposed, not hidden. But also different
from Shim6 in the sense that it operated at layer 3, and from what I
understand your goal was to actually develop new transport layers that
would be capable of taking the advantage of such exposed multihoming.
FWIW, imho whether the actual path selection is made on the transport
layer using different addresses or doing it at the ip layer, is an
implementation issue, but the main concepts apply in both cases. I
mean, the idea is to actually split the load in multiple paths, so the
usage of the network is more optimal and congestion is automatically
avoided
I suspect that load balancing of a single connection over multiple
paths
is a tough problem. How much research has there been on that, and do
we
have results that would already be usable for something like that?
I.e.,
is this solvable in the near future?
there is a lot of research already been done on this area actually and
the results are very interesting.
Most of this work was initiated by Kelly and it seems that the
theoretical part is well advanced so we could actually try to use some
of these results.
Some basic papers on this (selected by Damon cc)
* "Stability of end-to-end algorithms for joint routing and rate
control", Kelly and Voice.
* "Fluid models of integrated traffic and multipath routing",
Key, Massoulie
* "Optimal flow control and routing in multi-path networks",
Wang, Palaniswami, Low
* "Optimal Congestion Control with Multipath Routing Using TCP-
FAST and a Variant of RIP", Mallada, Paganini
* "Multi-path TCP: a joint congestion control and routing scheme
to exploit path diversity in the internet", Han, Shakkottai, Hollot,
Srikant, Towsley
* "Congestion control with adaptive multipath routing based on
optimization", Paganini
In any case, what I find interesting in this space is the different
design tradeoffs. A routing system that hides multihoming and provider
independence from the endpoints is easy for the endpoints and edge
networks. I.e., you do not have to change hosts in any way, every
network has a single prefix, renumbering is not necessary,
providers/network owners are in control of what kind of multihoming
and
TE is going on, etc.
the interesting part here is that TE can be performed in a different
way, but seems that both parties (the ISP and the users)still have
knobs to do this
regards, marcelo
But it also makes the routing system more expensive, because it has to
maintain a lot of information. Many of the RRG people are searching
for
a better organization of this information so that its maintenance
would
be cheaper -- but you are actually looking at removing some of this
information. I guess the main question is, can we substantially reduce
the costs of the routing system while keeping the same amount of
information and functionality in it? I'm not sure I know the answer
yet.
Another drawback of the hiding approach is that it might be ultimately
less capable, if you consider things like hosts being able to react on
transport layer timescales to congestion and their own communication
demands.
Jari
--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg
--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg