Mailing List Archive: Cisco: NSP

Experiences with l2tpv3/xconnect?

Index | Next | Previous | View Threaded

adrian at creative

Nov 4, 2009, 10:22 PM

Post #1 of 14 (254 views)
Permalink

Experiences with l2tpv3/xconnect?

G'day,

I've been asked by a customer to solve an L2 ethernet problem
and I'm investigating simply tunneling the required VLANs over
L2TPv3/xconnect.

Does anyone have any rough throughput (PPS in particular) info
they'd like to share ? And any other deployment info - actually,
in particular I'd like to know about fragmentation related issues.

I'm looking at the Cisco 28xx series (potentially the Cisco 2811)
but I'm concerned about hitting throughput ceilings.

Thanks,

Adrian

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

berghauz at gmail

Nov 4, 2009, 11:51 PM

Post #2 of 14 (242 views)
Permalink

Re: Experiences with l2tpv3/xconnect? [In reply to]

L2tpV3 very useful feature, but cause very high load on CPU on reciever
side.
On 1841 10Mbit/s xconnect channel cause near 40% CPU load.
2Mbit/s channel load CPU near 10%.
Max. throughput on 1841 without shaping 28Mbit/s(FULL CPU load).

WBR Aleksey Polyakoff ICQ:9001016
Ted Turner <http://www.brainyquote.com/quotes/authors/t/ted_turner.html> -
"Sports is like a war without the killing."

2009/11/5 Adrian Chadd <adrian[at]creative.net.au>

> G'day,
>
> I've been asked by a customer to solve an L2 ethernet problem
> and I'm investigating simply tunneling the required VLANs over
> L2TPv3/xconnect.
>
> Does anyone have any rough throughput (PPS in particular) info
> they'd like to share ? And any other deployment info - actually,
> in particular I'd like to know about fragmentation related issues.
>
> I'm looking at the Cisco 28xx series (potentially the Cisco 2811)
> but I'm concerned about hitting throughput ceilings.
>
> Thanks,
>
>
> Adrian
>
> _______________________________________________
> cisco-nsp mailing list cisco-nsp[at]puck.nether.net
> 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.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

rens at autempspourmoi

Nov 5, 2009, 12:08 AM

Post #3 of 14 (239 views)
Permalink

Re: Experiences with l2tpv3/xconnect? [In reply to]

You need to raise your MTU and the CPU load will go down.

PS: I'm not sure which IOS version supports baby giant frames on 1841, not
all do.

-----Original Message-----
From: cisco-nsp-bounces[at]puck.nether.net
[mailto:cisco-nsp-bounces[at]puck.nether.net] On Behalf Of Alexey Polyakov
Sent: jeudi 5 novembre 2009 8:52
To: Adrian Chadd
Cc: cisco-nsp[at]puck.nether.net
Subject: Re: [c-nsp] Experiences with l2tpv3/xconnect?

L2tpV3 very useful feature, but cause very high load on CPU on reciever
side.
On 1841 10Mbit/s xconnect channel cause near 40% CPU load.
2Mbit/s channel load CPU near 10%.
Max. throughput on 1841 without shaping 28Mbit/s(FULL CPU load).

WBR Aleksey Polyakoff ICQ:9001016
Ted Turner <http://www.brainyquote.com/quotes/authors/t/ted_turner.html> -
"Sports is like a war without the killing."

2009/11/5 Adrian Chadd <adrian[at]creative.net.au>

> G'day,
>
> I've been asked by a customer to solve an L2 ethernet problem
> and I'm investigating simply tunneling the required VLANs over
> L2TPv3/xconnect.
>
> Does anyone have any rough throughput (PPS in particular) info
> they'd like to share ? And any other deployment info - actually,
> in particular I'd like to know about fragmentation related issues.
>
> I'm looking at the Cisco 28xx series (potentially the Cisco 2811)
> but I'm concerned about hitting throughput ceilings.
>
> Thanks,
>
>
> Adrian
>
> _______________________________________________
> cisco-nsp mailing list cisco-nsp[at]puck.nether.net
> 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.nether.net
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.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

rens at autempspourmoi

Nov 5, 2009, 12:12 AM

Post #4 of 14 (239 views)
Permalink

Re: Experiences with l2tpv3/xconnect? [In reply to]

I have already done up to 400 Mbps with 2811 or 2821 (don't remember)
You just have to make sure your MTU is high enough depending on the frame
sizes you want to tunnel.

-----Original Message-----
From: cisco-nsp-bounces[at]puck.nether.net
[mailto:cisco-nsp-bounces[at]puck.nether.net] On Behalf Of Adrian Chadd
Sent: jeudi 5 novembre 2009 7:22
To: cisco-nsp[at]puck.nether.net
Subject: [c-nsp] Experiences with l2tpv3/xconnect?

G'day,

I've been asked by a customer to solve an L2 ethernet problem
and I'm investigating simply tunneling the required VLANs over
L2TPv3/xconnect.

Does anyone have any rough throughput (PPS in particular) info
they'd like to share ? And any other deployment info - actually,
in particular I'd like to know about fragmentation related issues.

I'm looking at the Cisco 28xx series (potentially the Cisco 2811)
but I'm concerned about hitting throughput ceilings.

Thanks,

Adrian

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

berghauz at gmail

Nov 5, 2009, 12:28 AM

Post #5 of 14 (240 views)
Permalink

Re: Experiences with l2tpv3/xconnect? [In reply to]

Becouse 1841 care only Fa interface, i think baby giant not supported at
all.
With 2811 the situation the same, because the Fa.

WBR Aleksey Polyakoff ICQ:9001016
Stephen Leacock<http://www.brainyquote.com/quotes/authors/s/stephen_leacock.html>
- "I detest life-insurance agents: they always argue that I shall some
day
die, which is not so."

2009/11/5 Rens <rens[at]autempspourmoi.be>

> You need to raise your MTU and the CPU load will go down.
>
> PS: I'm not sure which IOS version supports baby giant frames on 1841, not
> all do.
>
>
_______________________________________________
cisco-nsp mailing list cisco-nsp[at]puck.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

adrian at creative

Nov 5, 2009, 2:44 AM

Post #6 of 14 (237 views)
Permalink

Re: Experiences with l2tpv3/xconnect? [In reply to]

On Thu, Nov 05, 2009, Rens wrote:
> I have already done up to 400 Mbps with 2811 or 2821 (don't remember)
> You just have to make sure your MTU is high enough depending on the frame
> sizes you want to tunnel.

What PPS was this with though?

I'm worried about VoIP/PABX traffic causing much more increased CPU.

I don't have the option to up the MTU; the supplied underlying circuit
is an L2 ethernet metro ethernet style service.

Adrian

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

dale.shaw+cisco-nsp at gmail

Nov 5, 2009, 11:41 AM

Post #7 of 14 (226 views)
Permalink

Re: Experiences with l2tpv3/xconnect? [In reply to]

Hi Adrian,

On Thu, Nov 5, 2009 at 9:44 PM, Adrian Chadd <adrian[at]creative.net.au> wrote:
>
> I don't have the option to up the MTU; the supplied underlying circuit
> is an L2 ethernet metro ethernet style service.

Do you know for sure that the carrier MTU doesn't have the headroom you need?

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

adrian at creative

Nov 5, 2009, 5:46 PM

Post #8 of 14 (223 views)
Permalink

Re: Experiences with l2tpv3/xconnect? [In reply to]

On Fri, Nov 06, 2009, Dale Shaw wrote:

> > I don't have the option to up the MTU; the supplied underlying circuit
> > is an L2 ethernet metro ethernet style service.
>
> Do you know for sure that the carrier MTU doesn't have the headroom you need?

I'm going to make that assumption in case it is either true now, or becomes
true later.

Adrian

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

adrian at creative

Nov 5, 2009, 6:26 PM

Post #9 of 14 (222 views)
Permalink

Re: Experiences with l2tpv3/xconnect? [In reply to]

On Thu, Nov 05, 2009, Rens wrote:
> I have already done up to 400 Mbps with 2811 or 2821 (don't remember)
> You just have to make sure your MTU is high enough depending on the frame
> sizes you want to tunnel.

Just out of morbid curiousity - so will the router terminating L2TPv3 actually
fragment and reassemble L2TPv3 frames as needed, or is it hoping another
upstream router will fragment as needed?

Adrian

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

tvarriale at comcast

Nov 5, 2009, 6:56 PM

Post #10 of 14 (222 views)
Permalink

Re: Experiences with l2tpv3/xconnect? [In reply to]

Surely you mean 40mbps or a different platform?

tv
----- Original Message -----
From: "Rens" <rens[at]autempspourmoi.be>
To: "'Adrian Chadd'" <adrian[at]creative.net.au>; <cisco-nsp[at]puck.nether.net>
Sent: Thursday, November 05, 2009 2:12 AM
Subject: Re: [c-nsp] Experiences with l2tpv3/xconnect?

>I have already done up to 400 Mbps with 2811 or 2821 (don't remember)
> You just have to make sure your MTU is high enough depending on the frame
> sizes you want to tunnel.
>
> -----Original Message-----
> From: cisco-nsp-bounces[at]puck.nether.net
> [mailto:cisco-nsp-bounces[at]puck.nether.net] On Behalf Of Adrian Chadd
> Sent: jeudi 5 novembre 2009 7:22
> To: cisco-nsp[at]puck.nether.net
> Subject: [c-nsp] Experiences with l2tpv3/xconnect?
>
> G'day,
>
> I've been asked by a customer to solve an L2 ethernet problem
> and I'm investigating simply tunneling the required VLANs over
> L2TPv3/xconnect.
>
> Does anyone have any rough throughput (PPS in particular) info
> they'd like to share ? And any other deployment info - actually,
> in particular I'd like to know about fragmentation related issues.
>
> I'm looking at the Cisco 28xx series (potentially the Cisco 2811)
> but I'm concerned about hitting throughput ceilings.
>
> Thanks,
>
>
> Adrian
>
> _______________________________________________
> cisco-nsp mailing list cisco-nsp[at]puck.nether.net
> 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.nether.net
> 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.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

rens at autempspourmoi

Nov 6, 2009, 12:36 AM

Post #11 of 14 (219 views)
Permalink

Re: Experiences with l2tpv3/xconnect? [In reply to]

Indeed, I looked at the wrong lab tests.

Max I got out of a 2811 was around 90Mbps (1024 bytes)

-----Original Message-----
From: cisco-nsp-bounces[at]puck.nether.net
[mailto:cisco-nsp-bounces[at]puck.nether.net] On Behalf Of Tony Varriale
Sent: vendredi 6 novembre 2009 3:57
To: cisco-nsp[at]puck.nether.net
Subject: Re: [c-nsp] Experiences with l2tpv3/xconnect?

Surely you mean 40mbps or a different platform?

tv
----- Original Message -----
From: "Rens" <rens[at]autempspourmoi.be>
To: "'Adrian Chadd'" <adrian[at]creative.net.au>; <cisco-nsp[at]puck.nether.net>
Sent: Thursday, November 05, 2009 2:12 AM
Subject: Re: [c-nsp] Experiences with l2tpv3/xconnect?

>I have already done up to 400 Mbps with 2811 or 2821 (don't remember)
> You just have to make sure your MTU is high enough depending on the frame
> sizes you want to tunnel.
>
> -----Original Message-----
> From: cisco-nsp-bounces[at]puck.nether.net
> [mailto:cisco-nsp-bounces[at]puck.nether.net] On Behalf Of Adrian Chadd
> Sent: jeudi 5 novembre 2009 7:22
> To: cisco-nsp[at]puck.nether.net
> Subject: [c-nsp] Experiences with l2tpv3/xconnect?
>
> G'day,
>
> I've been asked by a customer to solve an L2 ethernet problem
> and I'm investigating simply tunneling the required VLANs over
> L2TPv3/xconnect.
>
> Does anyone have any rough throughput (PPS in particular) info
> they'd like to share ? And any other deployment info - actually,
> in particular I'd like to know about fragmentation related issues.
>
> I'm looking at the Cisco 28xx series (potentially the Cisco 2811)
> but I'm concerned about hitting throughput ceilings.
>
> Thanks,
>
>
> Adrian
>
> _______________________________________________
> cisco-nsp mailing list cisco-nsp[at]puck.nether.net
> 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.nether.net
> 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.nether.net
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.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

hl at r-kom

Nov 6, 2009, 1:22 AM

Post #12 of 14 (216 views)
Permalink

Re: Experiences with l2tpv3/xconnect? [In reply to]

On 06.11.09 10:26, Adrian Chadd wrote:
> On Thu, Nov 05, 2009, Rens wrote:
Hi,
> > I have already done up to 400 Mbps with 2811 or 2821 (don't remember)
> > You just have to make sure your MTU is high enough depending on the frame
> > sizes you want to tunnel.
I get the 2801 to tunnel about 40mbit, depending on package size of course.
I think 400mbit is more than possible.
>
> Just out of morbid curiousity - so will the router terminating L2TPv3 actually
> fragment and reassemble L2TPv3 frames as needed, or is it hoping another
> upstream router will fragment as needed?
Yes, both l2tp router will fragment and reassemble as needed, but you might
get problems if any transit router is fragmenting again. You should lower
your mtu at the external interface to prevent that. Furthermore you
will notice a big performance hit with packets causing fragmentation.

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

gert at greenie

Nov 6, 2009, 1:53 AM

Post #13 of 14 (217 views)
Permalink

Re: Experiences with l2tpv3/xconnect? [In reply to]

Hi,

On Fri, Nov 06, 2009 at 10:26:32AM +0800, Adrian Chadd wrote:
> Just out of morbid curiousity - so will the router terminating L2TPv3 actually
> fragment and reassemble L2TPv3 frames as needed, or is it hoping another
> upstream router will fragment as needed?

Well, as always this depends on "who is hitting the MTU wall" - if the
encapsulating router already knows "can't send this packet", it will
fragment itself, otherwise, a router on the path needs to do so.

Reassembly is always done on the receiving L2TPv3 router, and is expensive.

gert
--
USENET is *not* the non-clickable part of WWW!
//www.muc.de/~gert/
Gert Doering - Munich, Germany gert[at]greenie.muc.de
fax: +49-89-35655025 gert[at]net.informatik.tu-muenchen.de

adrian at creative

Nov 6, 2009, 2:51 AM

Post #14 of 14 (216 views)
Permalink

Re: Experiences with l2tpv3/xconnect? [In reply to]

On Fri, Nov 06, 2009, Gert Doering wrote:
> Hi,
>
> On Fri, Nov 06, 2009 at 10:26:32AM +0800, Adrian Chadd wrote:
> > Just out of morbid curiousity - so will the router terminating L2TPv3 actually
> > fragment and reassemble L2TPv3 frames as needed, or is it hoping another
> > upstream router will fragment as needed?
>
> Well, as always this depends on "who is hitting the MTU wall" - if the
> encapsulating router already knows "can't send this packet", it will
> fragment itself, otherwise, a router on the path needs to do so.
>
> Reassembly is always done on the receiving L2TPv3 router, and is expensive.

Absolutely. I just think I'm going to have to bite that.

I'll do up some basic testing and report back numbers once it is deployed.

Thanks,

Adrian

_______________________________________________
cisco-nsp mailing list cisco-nsp[at]puck.nether.net
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 lists@gossamer-threads.com