Re: [mpls] draft-fang-mpls-label-forwarding-no-swap - how much does it really save?

[Shahram Davari <[e-mail]>]

Hi Eric,

The new swapped labels (The outgoing  label that replaces the incoming label) need to be stored in a table. Using this draft reduces the number of swapped labels that needs to be store, regardless of  implementation. Don't you agree?


Thanks
Shahram

-----Original Message-----
From: Eric C Rosen [mailto:[e-mail]] 
Sent: Monday, July 20, 2015 10:20 AM
To: Shahram Davari; [e-mail]; [e-mail]
Cc: [e-mail]
Subject: Re: [mpls] draft-fang-mpls-label-forwarding-no-swap - how much does it really save?

On 7/20/2015 12:27 PM, Shahram Davari wrote:
> This draft actually saves a lot of memory as it requires only very
> little MPLS entry table and associated processing. There are only a
> few Ethernet MAC headers used in MPLS forwarding, typically one MAC
> per port, but there can be thousands Of swapped labels. So it is not
> a matter of 14 bytes vs 18 bytes.

This seems like purely a local implementation matter.  You still have to 
look up the label, don't you?  If you want to optimize for the case 
where the top label doesn't actually change its value, just use a 
14-byte rewrite string instead of an 18-byte string for those labels, 
and allow multiple label lookups to result in a pointer to the same 
14-byte rewrite string.

I don't really see the point of the draft, as this optimization does not 
require any architecture changes or any protocol changes.