[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: comments on draft-ietf-shim6-failure-detection-06.txt {2}



Hi Iljitsch, all,
Thank you very much!
Please see my comeents in line... :-)

Iljitsch van Beijnum schrieb:

On 30-sep-2006, at 16:47, Deguang Le wrote:

This paragraph presents "SHIM6 needs to avoid sending packets
concurrently over multiple path".


However, in RFC 3582 - Goals for IPv6 Site-Multihoming Architectures,
Load sharing is one of the goals of multihoming.
3.1.2.  Load Sharing
   By multihoming, a site should be able to distribute both inbound  and
   outbound traffic between multiple transit providers.  This goal is
   for concurrent use of the multiple transit providers, not just the
   usage of one provider over one interval of time and another  provider
   over a different interval.


Therfore, I would like to know if this sentence "SHIM6 needs to avoid
sending packets concurrently over multiple path" means that SHIM6
doesn't support load sharing?


The text you cite pertains to individual shim contexts. Often, a shim context will be used for a single TCP session, and it's considered suboptimal to have packets belonging to the same TCP session travel through different paths, because this introduces packet reordering which impedes TCP performance.

However, shim6 still supports load balancing by allowing different contexts to use different addresses, so that part of the contexts will generate traffic over one path and part of the contexts over another path.

Note though, that most traffic will be associated with contexts for which no shim6 address rewriting is active, so load balancing greatly depends on things outside the scope of shim6, such as information in the DNS and routing.

I agree with you.
Based on you comments, in order to avoid the confision, do you think it is more precise to revise the follwoing sentence? :-)

-Page8
   SHIM6 needs to avoid sending packets concurrently over multiple
   paths, ...

to

-Page8
SHIM6 needs to avoid sending packets, which belong to the same transport connection (or the same context?), concurrently over multiple
   paths, ...

Cheers,
Deguang