Mailing List Archive: Cisco: NSP

ME3600X architecture

Index | Next | Previous | View Threaded

bacon at walleyesoftware

Mar 29, 2012, 11:37 AM

Post #1 of 6 (865 views)
Permalink

ME3600X architecture

Primarily a question for Waris probably:

Is there anything out there about the "Carrier Ethernet ASIC" or the overall architecture of the box?

Looking at the board itself, it looks as though there's two separate but identical ASIC complexes on the board. My instinct says that the 10G ports are run by one ASIC and the 24 1G ports are run by the other ASIC, with some sort of bridge or bus between 'em.

I only just got my hands on one (and yes of course I ripped it apart) so I'm only just starting to answer questions - but there's so many nice docs on the 6500 and 4900s that I'm spoiled...

Thanks!
-bacon

_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

waris at cisco

Mar 31, 2012, 3:42 PM

Post #2 of 6 (808 views)
Permalink

Re: ME3600X architecture [In reply to]

Hi,

The architecture details are currently not available on CCO.

ME3800X/ME3600X has Cisco Carrier Ethernet ASIC. It is a Cisco ASIC and
it has all the features (QOS, MPLS, EVC) hardwired. The forwarding is
TCAM based hence NO performance impact upon enabling multiple features
at the same time. Scale advertised is multidimensional meaning number of
IPv4 routes and IPv6 can coexist at the same time since every feature
has its own TCAM space reserved. Push/pop/swap of MPLS labels won't have
any performance impact and same is true for QoS policies.

There are two Cisco Carrier Ethernet ASICs on the platform, one for the
24x1Gig and another one for the 2x10Gig SFP+ ports (you were right about
the ASIC distribution). There is a non blocking connectivity between the
two ASICs. 10Gig SFP+ ports also support 1xGig Fiber SFP without
license.

Multicast is done in the egress path and has no performance impact.

-Waris

-----Original Message-----
From: cisco-nsp-bounces [at] puck
[mailto:cisco-nsp-bounces [at] puck] On Behalf Of Jeff Bacon
Sent: Thursday, March 29, 2012 11:37 AM
To: cisco-nsp [at] puck
Subject: [c-nsp] ME3600X architecture

Primarily a question for Waris probably:

Is there anything out there about the "Carrier Ethernet ASIC" or the
overall architecture of the box?

Looking at the board itself, it looks as though there's two separate but
identical ASIC complexes on the board. My instinct says that the 10G
ports are run by one ASIC and the 24 1G ports are run by the other ASIC,
with some sort of bridge or bus between 'em.

I only just got my hands on one (and yes of course I ripped it apart) so
I'm only just starting to answer questions - but there's so many nice
docs on the 6500 and 4900s that I'm spoiled...

Thanks!

-bacon

_______________________________________________

cisco-nsp mailing list cisco-nsp [at] puck
<mailto:cisco-nsp [at] puck>
https://puck.nether.net/mailman/listinfo/cisco-nsp
<https://puck.nether.net/mailman/listinfo/cisco-nsp>

archive at http://puck.nether.net/pipermail/cisco-nsp/
<http://puck.nether.net/pipermail/cisco-nsp/>

_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

jstuxuhu0816 at gmail

Apr 2, 2012, 8:28 AM

Post #3 of 6 (797 views)
Permalink

Re: ME3600X architecture [In reply to]

Hi, Waris,
I am very interested about the 10g port license, I don't quite understand why the port need license, last time I used that port trouble shooting so much time, at last find that it need license.
Since I never see any other platform need license for the port use.

Thanks and regards,
Xu Hu

On 1 Apr, 2012, at 6:42, "Waris Sagheer (waris)" <waris [at] cisco> wrote:

> Hi,
>
> The architecture details are currently not available on CCO.
>
> ME3800X/ME3600X has Cisco Carrier Ethernet ASIC. It is a Cisco ASIC and
> it has all the features (QOS, MPLS, EVC) hardwired. The forwarding is
> TCAM based hence NO performance impact upon enabling multiple features
> at the same time. Scale advertised is multidimensional meaning number of
> IPv4 routes and IPv6 can coexist at the same time since every feature
> has its own TCAM space reserved. Push/pop/swap of MPLS labels won't have
> any performance impact and same is true for QoS policies.
>
> There are two Cisco Carrier Ethernet ASICs on the platform, one for the
> 24x1Gig and another one for the 2x10Gig SFP+ ports (you were right about
> the ASIC distribution). There is a non blocking connectivity between the
> two ASICs. 10Gig SFP+ ports also support 1xGig Fiber SFP without
> license.
>
> Multicast is done in the egress path and has no performance impact.
>
>
>
> -Waris
>
>
>
> -----Original Message-----
> From: cisco-nsp-bounces [at] puck
> [mailto:cisco-nsp-bounces [at] puck] On Behalf Of Jeff Bacon
> Sent: Thursday, March 29, 2012 11:37 AM
> To: cisco-nsp [at] puck
> Subject: [c-nsp] ME3600X architecture
>
>
>
> Primarily a question for Waris probably:
>
>
>
> Is there anything out there about the "Carrier Ethernet ASIC" or the
> overall architecture of the box?
>
>
>
> Looking at the board itself, it looks as though there's two separate but
> identical ASIC complexes on the board. My instinct says that the 10G
> ports are run by one ASIC and the 24 1G ports are run by the other ASIC,
> with some sort of bridge or bus between 'em.
>
>
>
> I only just got my hands on one (and yes of course I ripped it apart) so
> I'm only just starting to answer questions - but there's so many nice
> docs on the 6500 and 4900s that I'm spoiled...
>
>
>
> Thanks!
>
> -bacon
>
>
>
> _______________________________________________
>
> cisco-nsp mailing list cisco-nsp [at] puck
> <mailto:cisco-nsp [at] puck>
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> <https://puck.nether.net/mailman/listinfo/cisco-nsp>
>
> archive at http://puck.nether.net/pipermail/cisco-nsp/
> <http://puck.nether.net/pipermail/cisco-nsp/>
>
> _______________________________________________
> cisco-nsp mailing list cisco-nsp [at] puck
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/

_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

nick at foobar

Apr 2, 2012, 9:22 AM

Post #4 of 6 (816 views)
Permalink

Re: ME3600X architecture [In reply to]

On 02/04/2012 17:28, Xu Hu wrote:
> I am very interested about the 10g port license, I don't quite
> understand why the port need license

The 3600x needs the license because Cisco wants more of your money.

Nick
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

bacon at walleyesoftware

Apr 5, 2012, 5:07 AM

Post #5 of 6 (753 views)
Permalink

Re: ME3600X architecture [In reply to]

Hello Waris -

Thank you for the insight. I am curious about how much latency there is
when applied as a 10G-only device (e.g. a L3 border device linking a 10G
LAN to a 10G WAN ring, thus all traffic on the same ASIC) vs the normal
intended flow (1G aggregation). I just don't have the time to test it
at the moment.

Rosen-MVPN has been brought up here as a "coming soon". Is the support
for GRE-encap in the ASIC already and just not coded for yet, or would
any putative Rosen-MVPN support need to use s/w-based encap - and would
it require a "recirc" a la 6500s?

Many thanks,
-bacon

> Hi,
> The architecture details are currently not available on CCO.
> ME3800X/ME3600X has Cisco Carrier Ethernet ASIC. It is a Cisco ASIC
> and it has all the features (QOS, MPLS, EVC) hardwired. The
> forwarding is TCAM based hence NO performance impact upon enabling
> multiple features at the same time. Scale advertised is
> multidimensional meaning number of IPv4 routes and IPv6 can coexist
> at the same time since every feature has its own TCAM space
> reserved. Push/pop/swap of MPLS labels won't have any performance
> impact and same is true for QoS policies.
> There are two Cisco Carrier Ethernet ASICs on the platform, one for
> the 24x1Gig and another one for the 2x10Gig SFP+ ports (you were
> right about the ASIC distribution). There is a non blocking
> connectivity between the two ASICs. 10Gig SFP+ ports also support
> 1xGig Fiber SFP without license.
> Multicast is done in the egress path and has no performance impact.
>
> -Waris

_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

waris at cisco

Apr 6, 2012, 3:09 AM

Post #6 of 6 (783 views)
Permalink

Re: ME3600X architecture [In reply to]

Bacon,
I'll unicast you the slides.
Regarding MVPN, it is supported in hardware. EFT image is available for
testing.

-Waris

-----Original Message-----
From: Jeff Bacon [mailto:bacon [at] walleyesoftware]
Sent: Thursday, April 05, 2012 5:08 AM
To: Waris Sagheer (waris)
Cc: cisco-nsp [at] puck
Subject: RE: [c-nsp] ME3600X architecture

Hello Waris -

Thank you for the insight. I am curious about how much latency there is
when applied as a 10G-only device (e.g. a L3 border device linking a 10G
LAN to a 10G WAN ring, thus all traffic on the same ASIC) vs the normal
intended flow (1G aggregation). I just don't have the time to test it at
the moment.

Rosen-MVPN has been brought up here as a "coming soon". Is the support
for GRE-encap in the ASIC already and just not coded for yet, or would
any putative Rosen-MVPN support need to use s/w-based encap - and would
it require a "recirc" a la 6500s?

Many thanks,
-bacon

> Hi,
> The architecture details are currently not available on CCO.
> ME3800X/ME3600X has Cisco Carrier Ethernet ASIC. It is a Cisco ASIC
> and it has all the features (QOS, MPLS, EVC) hardwired. The forwarding

> is TCAM based hence NO performance impact upon enabling multiple
> features at the same time. Scale advertised is multidimensional
> meaning number of IPv4 routes and IPv6 can coexist at the same time
> since every feature has its own TCAM space reserved. Push/pop/swap of
> MPLS labels won't have any performance impact and same is true for QoS

> policies.
> There are two Cisco Carrier Ethernet ASICs on the platform, one for
> the 24x1Gig and another one for the 2x10Gig SFP+ ports (you were right

> about the ASIC distribution). There is a non blocking connectivity
> between the two ASICs. 10Gig SFP+ ports also support 1xGig Fiber SFP
> without license.
> Multicast is done in the egress path and has no performance impact.
>
> -Waris

_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

Index | Next | Previous | View Threaded

Interested in having your list archived? Contact Gossamer Threads