[rbak-nsp] forwarding policy

[Ian Calderbank]

Redback does forwarding using the PPA, end of story. (apart from some very
special exceptions).

Process-switching concept doesn't exist outside of cisco. Seems your
friendly local redback SE isn't giving you the basic architecture lesson?

<history>
Ye olde Cisco way of  forwarding packets = process switching, fast-switching
yada yada
Cisco invented Cef then better DCEF = Distributed forwarding using line card
processors (VIP's on 7500's in the mid 90s) to perform all forwarding
actions (except those that CEF didn't support.. which was the issue).

Juniper did it properly with a box with a "strict separation of control and
forwarding planes" architecture (enter stage left to thunderous applause the
M series). impossible for a transit packet to touch the control processor.
The forwarding processor (either a shared forwarding blade, or a chip on a
line card) does all the forwarding work. Control cpu cruises at 1% if the
network is converged/stable, regardless of the forwarding load.

Redback followed suit with the SE. Did a bit more work on modularising the
control plane than J, different backplane design, different asics of course,
but still the fundamental separation of forwarding and control processors
and data paths, transit packet unable to touch the RP.
</history>

As david mentions there are corner case applications where you might
deliberately redirect the packet to the RP such as http url redirect .
Session Border controller does the same for SIP I would expect.

For a regular forwarding policy, it just adds another link in the PPA's
chain of actions (called a PED graph) that are to be applied to that
circuit. Each PED costs you a bit of resource (mem , cpu) on the PPA. But so
long as your PPA's are not overcooked - don't worry about it.

cheers
Ian

-----Original Message-----
From: redback-nsp-bounces at puck.nether.net
[mailto:redback-nsp-bounces at puck.nether.net] On Behalf Of
redback-nsp-request at puck.nether.net
Sent: 28 February 2011 17:00
To: redback-nsp at puck.nether.net
Subject: redback-nsp Digest, Vol 38, Issue 13

Send redback-nsp mailing list submissions to
	redback-nsp at puck.nether.net

To subscribe or unsubscribe via the World Wide Web, visit
	https://puck.nether.net/mailman/listinfo/redback-nsp
or, via email, send a message with subject or body 'help' to
	redback-nsp-request at puck.nether.net

You can reach the person managing the list at
	redback-nsp-owner at puck.nether.net

When replying, please edit your Subject line so it is more specific than
"Re: Contents of redback-nsp digest..."


Today's Topics:

   1. forwarding policy (Richard Clayton)
   2. Re: forwarding policy (david.freedman at uk.clara.net)


----------------------------------------------------------------------

Message: 1
Date: Sun, 27 Feb 2011 22:05:43 +0000
From: Richard Clayton <sledge121 at gmail.com>
To: redback-nsp at puck.nether.net
Subject: [rbak-nsp] forwarding policy
Message-ID:
	<AANLkTinwxkUwkiStm6PuGt=xdPaf1TjPNeeyid7f=K-c at mail.gmail.com>
Content-Type: text/plain; charset="iso-8859-1"

Am I correct in thinking that a Redback forwarding policy is process
switched, the same as Cisco policy-based routing.

Thanks
Rick
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<https://puck.nether.net/pipermail/redback-nsp/attachments/20110227/a50d24e4
/attachment-0001.html>

------------------------------

Message: 2
Date: Sun, 27 Feb 2011 22:10:28 +0000 (GMT)
From: david.freedman at uk.clara.net
To: sledge121 at gmail.com
Cc: redback-nsp at puck.nether.net
Subject: Re: [rbak-nsp] forwarding policy
Message-ID:
	<Pine.BSF.4.58.1102272206340.31883 at singularity.convergence.cx>
Content-Type: TEXT/PLAIN; charset=US-ASCII

>Am I correct in thinking that a Redback forwarding policy is process 
>switched

Not quite sure what you mean here, my understanding is that unless you are
using XCRP services (such as the http server), then the forwarding actions
occur in hardware, on the PPA,


[thingy]redback# show forward policy foo

Policy-Name                Type     Grid    Qs Slots  Ports  Bound  DnLd
Status
foo                        forward  14      0  1      4      in

Slot#:       1   2   3   4   5   6   7   8   9  10  11  12  13  14
iPPA dnld:
ePPA dnld:
iPPA ports:  4
ePPA ports:

Class-Name          Action  Mode  IP-Addr/Option   Bound  Int,msec
Output-Name
redirect-class      redir   local
drop-class          drop

Total policy map: 1

I believe this is possible because these are FPGA , though I'm sure there
are some internal limits to how nested operations can be.

Dave.



------------------------------

_______________________________________________
redback-nsp mailing list
redback-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/redback-nsp


End of redback-nsp Digest, Vol 38, Issue 13
*******************************************