Hi Noel,

> -----Original Message-----
> From: Noel Chiappa [mailto:[email protected]]
> Sent: Thursday, May 30, 2013 9:44 AM
> To: [email protected]
> Cc: [email protected]
> Subject: Re: [lisp] AD Evaluation: draft-ietf-lisp-deployment
> 
>     > From: "Templin, Fred L" <[email protected]>
> 
>     >> Can you please give an example of a tunnel where the
> encapsulation
>     >> overhead does not result in a decrease of the E2E PMTU?
> 
>     > SEAL is an example of a tunnel type where the added overhead does
> not
>     > decrease the effective end-to-end path MTU for MTUs up to 1500
> bytes.
>     > Another example is CAPWAP.
> 
> Both SEAL and CAPWAP apparently rely on fragmentation/reassembly to do
> this.

Both SEAL and CAPWAP have their own segmentation/reassembly
schemes that are apart from IP fragmentation and do not share
some of the same pitfalls.

> The last time we tried to use fragmentation/reassembly on a large
> scale,
> it just didn't work:

I agree that reassembly in an egress router that is located somewhere
in the middle of a large network doesn't scale. For egress routers
that are close to the destination end systems, however, the scaling
is bounded. In general, the closer the egress router is to the end
systems the more practical reassembly becomes - especially if the
egress router and end system are one and the same.  

>   Fragmentation Considered Harmful
>   ftp://ftp.hpl.hp.com/pub/dec/WRL/research-reports/WRL-TR-87.3.ps

SEAL (and CAPWAP too I assume) took this work as guidance on how
to avoid the pitfalls.

> In light of that, what _experience_ is there with these protocols to
> show
> that fragmentation/reassembly is a _practical_ approach?

We are using SEAL internally within Boeing. And, my understanding
is that CAPWAP is widely used.
 
> (Since CAPWAP supports and MTU somewhat larger than jumbograms, it's
> not
> clear if they have much actual experience with
> fragmentation/reassembly.)

CAPWAP assures a tunnel MTU of at least 1500 bytes and uses
segmentation/reassembly to accomplish this when necessary.
But, that is not the same thing as IP fragmentation/reassembly.

Thanks - Fred
 
>       Noel
_______________________________________________
lisp mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/lisp

Reply via email to