Re: [trill] Forwarding table for TRILL

Muhammad Mohsin Sardar <mohsin.sardar@seecs.edu.pk> Fri, 24 August 2012 17:20 UTC

MIME-Version: 1.0
In-Reply-To: <CAF4+nEHSjHd8x6y_ZX+xS3PBS3TNbSgs6xOfLU+GjvXN9qUrOQ@mail.gmail.com>
References: <CAEfTPhXsKWVygwjnDabaY_ZihghmF_V7xdhpgKD8vp9j67acxQ@mail.gmail.com> <CAEfTPhVGVhUk0m4yUFV2xTnHxGC2vDFWpXN7qhMhtk-dSwTpEA@mail.gmail.com> <CAF4+nEHSjHd8x6y_ZX+xS3PBS3TNbSgs6xOfLU+GjvXN9qUrOQ@mail.gmail.com>
From: Muhammad Mohsin Sardar <mohsin.sardar@seecs.edu.pk>
Date: Fri, 24 Aug 2012 22:19:37 +0500
Message-ID: <CAEfTPhUiCGw29kaRYjnJOjevR1TbGW1Q4442Fnzm10Op9-k1pw@mail.gmail.com>
To: Donald Eastlake <d3e3e3@gmail.com>
Content-Type: multipart/alternative; boundary="047d7b34352667a4ba04c8062f38"
Cc: trill@ietf.org
Subject: Re: [trill] Forwarding table for TRILL
Precedence: list

Hi Donald,
Thanks for reply!
Please clarify inline question.

On Fri, Aug 24, 2012 at 5:58 PM, Donald Eastlake <d3e3e3@gmail.com> wrote:

> Hi Muhammad,
>
> On Fri, Aug 24, 2012 at 2:11 AM, Muhammad Mohsin Sardar
> <mohsin.sardar@seecs.edu.pk> wrote:
> > Hi,
> >
> > As per my understanding from RFC6325 that TRILL has two forwarding
> > tables, one having the record of nicknames for RBridges and other
> > haveing the MAC addresses for end stations.
> >
> > Intermediate RBridges only have nicknames table to forward encapsulated
> > packets.
> >
> > Whereas ingress/egress RBridges have both nickname and MAC Address table.
>
> An implementer can structure their tables however they want as long as
> the bits on the wire and the external behavior is the same; however,
> having two tables as you describe, one used by the native frame
> ingress/egress function and one used by the TRILL Data forwarding
> function is a reasonable way to look at it.
>
> > Question:
> > Does ingress/egress RBridges need a forwarding table having the size
> > equal to number of end nodes(for MAC) + total RBridges(for
> > Nickname)?
>
> Sort of.
>
> The ingress lookup is by VLAN+MAC to get an egress nickname (or find
> that it is not in the table) while the egress lookup is by VLAN+MAC to
> get a port (or find that it is not in the table). There is, of course,
> no requirement that the ingress/egress table(s) be exactly "equal" to
> the number of end station in the set of VLANs that the RBridge is
> interested in and that is probably almost never true.
>
Do we need ingress/egress table(s) nearly "equal" to the number of end
stations?
If we need MAC to nickname mapping for all end stations, Do we need a table
size nearly equal to the number of end stations?
Therefore, a larger ingress/egress table is required in case of data center
networks of hundred thousand end servers.

>
> On the other hand, there is a requirement that an RBridge be able to
> hold the link state which includes the nicknames of all other RBridges
> (otherwise it is in overflow, see draft-ietf-trill-clear-correct) and
> be able to look up the egress nickname from a TRILL Data frame to
> determine which port (unicast) or ports (multi-destinaiton) to forward
> the TRILL Data frame out of.
>
> Thanks,
> Donald
> =============================
>  Donald E. Eastlake 3rd   +1-508-333-2270 (cell)
>  155 Beaver Street, Milford, MA 01757 USA
>  d3e3e3@gmail.com
>
> > Thanks!
> >
> > --
> > Best Regards,
> > Muhammad Mohsin Sardar
>



-- 
Best Regards,
Muhammad Mohsin Sardar

[trill] Forwarding table for TRILL Muhammad Mohsin Sardar
Re: [trill] Forwarding table for TRILL Donald Eastlake
Re: [trill] Forwarding table for TRILL Muhammad Mohsin Sardar
Re: [trill] Forwarding table for TRILL Joe Touch