Mailing List Archive

Nope, just very little chatter.

BR.



> Le 28 janv. 2019 à 20:32, joe mcguckin <joe@via.net> a écrit :
>
> Is there a new list for Foundry/Brocade/Extreme?
>
> Thanks,
>
> joe
>
>
> Joe McGuckin
> ViaNet Communications
>
> joe@via.net
> 650-207-0372 cell
> 650-213-1302 office
> 650-969-2124 fax
>
>
>
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

I think everyone is phasing out their X-scalability hardware in anticipation of the global routing table growing larger than 786k IPv4 routes. These boxes don't have a long life ahead of them as a DFZ router, unless you only use X2-scalability cards in your chassis.. and with operators essentially being forced to upgrade to a completely different platform (SLX or $other_vendor), Brextreme's advantage of being the incumbent vendor in the network is pretty much completely negated.

Finally, as a BDA, let me tell you that being stabbed in the back when it comes to my certification record has not benefited my opinion of the new owner(s) of the Brocade diaspora. Why they decided to not only pull the plug on the entire Brocade education program, but also delete the historical certification database from CertMetrics, is beyond me. It has become impossible to check the validity of years' worth of Brocade credentials.

Best regards,
Martijn

On 28 January 2019 21:35:56 CET, Youssef Bengelloun-Zahr <bengelly@gmail.com> wrote:

Nope, just very little chatter.

BR.



Le 28 janv. 2019 à 20:32, joe mcguckin <joe@via.net> a écrit :

Is there a new list for Foundry/Brocade/Extreme?

Thanks,

joe


Joe McGuckin
ViaNet Communications

joe@via.net
650-207-0372 cell
650-213-1302 office
650-969-2124 fax
________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp
________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

--
Sent from my Android device with K-9 Mail. Please excuse my brevity.

We‘re testing the SLX for a while now but figured that quite a bit of the wheel got reinvented, or, to be more precise, the new wheel is only halfway round as of today.

Any suggestions for $other_vendor? I actually liked the hardware of the SLX - 24*10G and 12*100G.

Meanwhile, BGP4 aggregation is our new buddy... #LifeBeyond750k.

--
Fredy Kuenzler
Init7 (Switzerland) Ltd.
Technoparkstrasse 5
CH-8406 Winterthur
Switzerland

http://www.init7.net/


> Am 28.01.2019 um 22:06 schrieb i3D.net - Martijn Schmidt <martijnschmidt@i3d.net>:
>
> I think everyone is phasing out their X-scalability hardware in anticipation of the global routing table growing larger than 786k IPv4 routes. These boxes don't have a long life ahead of them as a DFZ router, unless you only use X2-scalability cards in your chassis.. and with operators essentially being forced to upgrade to a completely different platform (SLX or $other_vendor), Brextreme's advantage of being the incumbent vendor in the network is pretty much completely negated.
>
> Finally, as a BDA, let me tell you that being stabbed in the back when it comes to my certification record has not benefited my opinion of the new owner(s) of the Brocade diaspora. Why they decided to not only pull the plug on the entire Brocade education program, but also delete the historical certification database from CertMetrics, is beyond me. It has become impossible to check the validity of years' worth of Brocade credentials.
>
> Best regards,
> Martijn
>
>> On 28 January 2019 21:35:56 CET, Youssef Bengelloun-Zahr <bengelly@gmail.com> wrote:
>> Nope, just very little chatter.
>>
>> BR.
>>
>>
>>
>>> Le 28 janv. 2019 à 20:32, joe mcguckin <joe@via.net> a écrit :
>>>
>>> Is there a new list for Foundry/Brocade/Extreme?
>>>
>>> Thanks,
>>>
>>> joe
>>>
>>>
>>> Joe McGuckin
>>> ViaNet Communications
>>>
>>> joe@via.net
>>> 650-207-0372 cell
>>> 650-213-1302 office
>>> 650-969-2124 fax
>>> foundry-nsp mailing list
>>> foundry-nsp@puck.nether.net
>>> http://puck.nether.net/mailman/listinfo/foundry-nsp
>> foundry-nsp mailing list
>> foundry-nsp@puck.nether.net
>> http://puck.nether.net/mailman/listinfo/foundry-nsp
>
> --
> Sent from my Android device with K-9 Mail. Please excuse my brevity.
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp

Hi,

I guess most of the users simply replaced their Brocade boxes with
Juniper, Cisco or Arista and have no need to use that list here anymore.
;)

kind regards
Rolf

> Is there a new list for Foundry/Brocade/Extreme?
>
> Thanks,
>
> joe
>
>
> Joe McGuckin
> ViaNet Communications
>

_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

Yeap... I think its quite a game over for the Foundry-Brocade history, since it has been sold to Broadcom/Extreme...

Le 28 janvier 2019 22:31:08 GMT+01:00, "Rolf Hanßen" <nsp@rhanssen.de> a écrit :
>Hi,
>
>I guess most of the users simply replaced their Brocade boxes with
>Juniper, Cisco or Arista and have no need to use that list here
>anymore.
>;)
>
>kind regards
>Rolf
>
>> Is there a new list for Foundry/Brocade/Extreme?
>>
>> Thanks,
>>
>> joe
>>
>>
>> Joe McGuckin
>> ViaNet Communications
>>
>
>_______________________________________________
>foundry-nsp mailing list
>foundry-nsp@puck.nether.net
>http://puck.nether.net/mailman/listinfo/foundry-nsp

--
Envoyé de mon appareil Android avec Courriel K-9 Mail. Veuillez excuser ma brièveté.

We went from Brocade RX/MLX to Juniper MX960s to Arista 7500Rs.

Liked the Brocade no-nonsense attitude, loved JunOS but needed Arista
density (36x100G) after a while. If your requirements are not too
complicated the 7500Rs do a great job.

Rens
YISP

On 1/28/19 10:31 PM, "Rolf Hanßen" wrote:
> Hi,
>
> I guess most of the users simply replaced their Brocade boxes with
> Juniper, Cisco or Arista and have no need to use that list here anymore.
> ;)
>
> kind regards
> Rolf
>
>> Is there a new list for Foundry/Brocade/Extreme?
>>
>> Thanks,
>>
>> joe
>>
>>
>> Joe McGuckin
>> ViaNet Communications
>>
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

If I recall correctly you're using a lot of 10G ports in the Init7 FTTO hubs right? Have you considered getting a box with lots of QSFP+/QSFP28 ports and doing some breakout with patchpanels? This can be done with QSFP+ MPO LR4 transceivers too since you probably need singlemode, the folks from FlexOptix can hook you up with those!

We are, by the way, moving to Juniper hardware in the DFZ routing layer ourselves. The MX204 is our new bread and butter.

Best regards,
Martijn

On 28 January 2019 22:21:28 CET, Fredy Kuenzler <kuenzler@init7.net> wrote:
We‘re testing the SLX for a while now but figured that quite a bit of the wheel got reinvented, or, to be more precise, the new wheel is only halfway round as of today.

Any suggestions for $other_vendor? I actually liked the hardware of the SLX - 24*10G and 12*100G.

Meanwhile, BGP4 aggregation is our new buddy... #LifeBeyond750k.

--
Fredy Kuenzler
Init7 (Switzerland) Ltd.
Technoparkstrasse 5
CH-8406 Winterthur
Switzerland

http://www.init7.net/


Am 28.01.2019 um 22:06 schrieb i3D.net<http://i3D.net> - Martijn Schmidt <martijnschmidt@i3d.net<mailto:martijnschmidt@i3d.net>>:

I think everyone is phasing out their X-scalability hardware in anticipation of the global routing table growing larger than 786k IPv4 routes. These boxes don't have a long life ahead of them as a DFZ router, unless you only use X2-scalability cards in your chassis.. and with operators essentially being forced to upgrade to a completely different platform (SLX or $other_vendor), Brextreme's advantage of being the incumbent vendor in the network is pretty much completely negated.

Finally, as a BDA, let me tell you that being stabbed in the back when it comes to my certification record has not benefited my opinion of the new owner(s) of the Brocade diaspora. Why they decided to not only pull the plug on the entire Brocade education program, but also delete the historical certification database from CertMetrics, is beyond me. It has become impossible to check the validity of years' worth of Brocade credentials.

Best regards,
Martijn

On 28 January 2019 21:35:56 CET, Youssef Bengelloun-Zahr <bengelly@gmail.com<mailto:bengelly@gmail.com>> wrote:

Nope, just very little chatter.

BR.



Le 28 janv. 2019 à 20:32, joe mcguckin <joe@via.net<mailto:joe@via.net>> a écrit :

Is there a new list for Foundry/Brocade/Extreme?

Thanks,

joe


Joe McGuckin
ViaNet Communications

joe@via.net<mailto:joe@via.net>
650-207-0372 cell
650-213-1302 office
650-969-2124 fax
________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net<mailto:foundry-nsp@puck.nether.net>
http://puck.nether.net/mailman/listinfo/foundry-nsp
________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net<mailto:foundry-nsp@puck.nether.net>
http://puck.nether.net/mailman/listinfo/foundry-nsp

--
Sent from my Android device with K-9 Mail. Please excuse my brevity.
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net<mailto:foundry-nsp@puck.nether.net>
http://puck.nether.net/mailman/listinfo/foundry-nsp

--
Sent from my Android device with K-9 Mail. Please excuse my brevity.

I'm actively moving customers to Juniper.

--

--
Regards,
--
Brant I. Stevens, Principal & Consulting Architect
branto@argentiumsolutions.com
http://argentiumsolutions.com <http://www.argentiumsolutions.com>

Fredy Kuenzler wrote on 1/28/19 4:21 PM:
> We‘re testing the SLX for a while now but figured that quite a bit of
> the wheel got reinvented, or, to be more precise, the new wheel is
> only halfway round as of today.
>
> Any suggestions for $other_vendor? I actually liked the hardware of
> the SLX - 24*10G and 12*100G.
>
> Meanwhile, BGP4 aggregation is our new buddy... #LifeBeyond750k.
>
> --
> Fredy Kuenzler
> Init7 (Switzerland) Ltd.
> Technoparkstrasse 5
> CH-8406 Winterthur
> Switzerland
>
> http://www.init7.net/
>
>
> Am 28.01.2019 um 22:06 schrieb i3D.net <http://i3D.net> - Martijn
> Schmidt <martijnschmidt@i3d.net <mailto:martijnschmidt@i3d.net>>:
>
>> I think everyone is phasing out their X-scalability hardware in
>> anticipation of the global routing table growing larger than 786k
>> IPv4 routes. These boxes don't have a long life ahead of them as a
>> DFZ router, unless you only use X2-scalability cards in your
>> chassis.. and with operators essentially being forced to upgrade to a
>> completely different platform (SLX or $other_vendor), Brextreme's
>> advantage of being the incumbent vendor in the network is pretty much
>> completely negated.
>>
>> Finally, as a BDA, let me tell you that being stabbed in the back
>> when it comes to my certification record has not benefited my opinion
>> of the new owner(s) of the Brocade diaspora. Why they decided to not
>> only pull the plug on the entire Brocade education program, but also
>> delete the historical certification database from CertMetrics, is
>> beyond me. It has become impossible to check the validity of years'
>> worth of Brocade credentials.
>>
>> Best regards,
>> Martijn
>>
>> On 28 January 2019 21:35:56 CET, Youssef Bengelloun-Zahr
>> <bengelly@gmail.com <mailto:bengelly@gmail.com>> wrote:
>>
>> Nope, just very little chatter.
>>
>> BR.
>>
>>
>>
>> Le 28 janv. 2019 à 20:32, joe mcguckin <joe@via.net
>> <mailto:joe@via.net>> a écrit : Is there a new list for
>> Foundry/Brocade/Extreme? Thanks, joe Joe McGuckin ViaNet
>> Communications joe@via.net <mailto:joe@via.net> 650-207-0372
>> cell 650-213-1302 office 650-969-2124 fax
>> ------------------------------------------------------------------------
>> foundry-nsp mailing list foundry-nsp@puck.nether.net
>> <mailto:foundry-nsp@puck.nether.net>
>> http://puck.nether.net/mailman/listinfo/foundry-nsp
>>
>> ------------------------------------------------------------------------
>> foundry-nsp mailing list
>> foundry-nsp@puck.nether.net <mailto:foundry-nsp@puck.nether.net>
>> http://puck.nether.net/mailman/listinfo/foundry-nsp
>>
>>
>> --
>> Sent from my Android device with K-9 Mail. Please excuse my brevity.
>> _______________________________________________
>> foundry-nsp mailing list
>> foundry-nsp@puck.nether.net <mailto:foundry-nsp@puck.nether.net>
>> http://puck.nether.net/mailman/listinfo/foundry-nsp
>
>
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp

That's probably a reason, also the X2 card pricing is not so attractive.
I also wonder when / if they will start a trade-in program for MLX to
SLX, but seems they missed this opportunity with existing clients
already. :(

Additionally the SLX9850/9540 got its 1.5M route update late (March
2018?) and the SLX9640 is kind of being late to the party, too (but
still a cool product!), so that we went for the SLX9540 for peering
locations (and they needed 3 months to ship … ).

For one point I thought that some kind of adapter card, that makes X2
cards work inside the SLX chassis, would protect a future investment and
also running the power-pc based NetIron-OS inside the KVM hypervisor
would be neat.

I also think that marketing or product management of the former Brocade
products are (purposely?) pretty weak and besides they are talking about
inside architecture or visibility (SLX) platform, the hardware products
are hardly visible or known at all. The last one-day sales presentation
I have attended a year after the transition, they didn't even mention
former "Brocade" products and talked about Extreme becoming THE
"software / analytics vendor" for generic network devices…

Well, what about being both? A software vendor and a hardware vendor
(with touch control… )? The Brocade buy-out was such a steal and then
they forgot to bring the customers into a safe harbor. Still to be
mentioned that the whole Broadcom-Brocade-Qualcomm-US-Government
disaster may had its own impact, too.

So what they really need is a momentum shift. Bring the SLX-line out on
the road where it matters and it can work, sponsor the right events,
understand why people currently don't choose the products and work on
this issues, show valid upgrade paths for current customers and develop
a strategy for not loosing more customers to other vendors, advance the
development (5 months for a critical bug?) and hand out the Brocade
strategy behind 2020.

Jörg


On 28 Jan 2019, at 22:06, i3D.net - Martijn Schmidt wrote:

> I think everyone is phasing out their X-scalability hardware in
> anticipation of the global routing table growing larger than 786k IPv4
> routes. These boxes don't have a long life ahead of them as a DFZ
> router, unless you only use X2-scalability cards in your chassis.. and
> with operators essentially being forced to upgrade to a completely
> different platform (SLX or $other_vendor), Brextreme's advantage of
> being the incumbent vendor in the network is pretty much completely
> negated.
>
> Finally, as a BDA, let me tell you that being stabbed in the back when
> it comes to my certification record has not benefited my opinion of
> the new owner(s) of the Brocade diaspora. Why they decided to not only
> pull the plug on the entire Brocade education program, but also delete
> the historical certification database from CertMetrics, is beyond me.
> It has become impossible to check the validity of years' worth of
> Brocade credentials.
>
> Best regards,
> Martijn
>
> On 28 January 2019 21:35:56 CET, Youssef Bengelloun-Zahr
> <bengelly@gmail.com> wrote:
>
> Nope, just very little chatter.
>
> BR.
>
>
>
> Le 28 janv. 2019 à 20:32, joe mcguckin <joe@via.net> a écrit :
>
> Is there a new list for Foundry/Brocade/Extreme?
>
> Thanks,
>
> joe
>
>
> Joe McGuckin
> ViaNet Communications
>
> joe@via.net
> 650-207-0372 cell
> 650-213-1302 office
> 650-969-2124 fax
> ________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp
> ________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp
>
> --
> Sent from my Android device with K-9 Mail. Please excuse my brevity.


> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp

There is also an active community:
https://community.extremenetworks.com/data-center-slx-vdx-mlx-ces-232983

On 28 Jan 2019, at 20:32, joe mcguckin wrote:

> Is there a new list for Foundry/Brocade/Extreme?
>
> Thanks,
>
> joe
>
>
> Joe McGuckin
> ViaNet Communications
>
> joe@via.net
> 650-207-0372 cell
> 650-213-1302 office
> 650-969-2124 fax
>
>
>
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

Yeah, except: that "update" is just a BGP compression algorithm on the SLX-OS software. Your hardware still only carries 1M routes, but they simply don't install overlapping subnets with the same next-hop into FIB. So you are really just stuck with hardware that has less route scalability than the good ol' X2-scalability MLXe linecards in every measurable way.

Best regards,
Martijn

On 1/29/19 10:31 AM, Jörg Kost wrote:

That's probably a reason, also the X2 card pricing is not so attractive. I also wonder when / if they will start a trade-in program for MLX to SLX, but seems they missed this opportunity with existing clients already. :(

Additionally the SLX9850/9540 got its 1.5M route update late (March 2018?) and the SLX9640 is kind of being late to the party, too (but still a cool product!), so that we went for the SLX9540 for peering locations (and they needed 3 months to ship … ).

For one point I thought that some kind of adapter card, that makes X2 cards work inside the SLX chassis, would protect a future investment and also running the power-pc based NetIron-OS inside the KVM hypervisor would be neat.

I also think that marketing or product management of the former Brocade products are (purposely?) pretty weak and besides they are talking about inside architecture or visibility (SLX) platform, the hardware products are hardly visible or known at all. The last one-day sales presentation I have attended a year after the transition, they didn't even mention former "Brocade" products and talked about Extreme becoming THE "software / analytics vendor" for generic network devices…

Well, what about being both? A software vendor and a hardware vendor (with touch control… )? The Brocade buy-out was such a steal and then they forgot to bring the customers into a safe harbor. Still to be mentioned that the whole Broadcom-Brocade-Qualcomm-US-Government disaster may had its own impact, too.

So what they really need is a momentum shift. Bring the SLX-line out on the road where it matters and it can work, sponsor the right events, understand why people currently don't choose the products and work on this issues, show valid upgrade paths for current customers and develop a strategy for not loosing more customers to other vendors, advance the development (5 months for a critical bug?) and hand out the Brocade strategy behind 2020.

Jörg


On 28 Jan 2019, at 22:06, i3D.net - Martijn Schmidt wrote:

I think everyone is phasing out their X-scalability hardware in anticipation of the global routing table growing larger than 786k IPv4 routes. These boxes don't have a long life ahead of them as a DFZ router, unless you only use X2-scalability cards in your chassis.. and with operators essentially being forced to upgrade to a completely different platform (SLX or $other_vendor), Brextreme's advantage of being the incumbent vendor in the network is pretty much completely negated.

Finally, as a BDA, let me tell you that being stabbed in the back when it comes to my certification record has not benefited my opinion of the new owner(s) of the Brocade diaspora. Why they decided to not only pull the plug on the entire Brocade education program, but also delete the historical certification database from CertMetrics, is beyond me. It has become impossible to check the validity of years' worth of Brocade credentials.

Best regards,
Martijn

On 28 January 2019 21:35:56 CET, Youssef Bengelloun-Zahr <bengelly@gmail.com><mailto:bengelly@gmail.com> wrote:

Nope, just very little chatter.

BR.




Le 28 janv. 2019 à 20:32, joe mcguckin <joe@via.net><mailto:joe@via.net> a écrit :

Is there a new list for Foundry/Brocade/Extreme?

Thanks,

joe


Joe McGuckin
ViaNet Communications

joe@via.net<mailto:joe@via.net>
650-207-0372 cell
650-213-1302 office
650-969-2124 fax
________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net<mailto:foundry-nsp@puck.nether.net>
http://puck.nether.net/mailman/listinfo/foundry-nsp

________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net<mailto:foundry-nsp@puck.nether.net>
http://puck.nether.net/mailman/listinfo/foundry-nsp


--
Sent from my Android device with K-9 Mail. Please excuse my brevity.

_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net<mailto:foundry-nsp@puck.nether.net>
http://puck.nether.net/mailman/listinfo/foundry-nsp

Would be interested to know your aggregate / filter criteria, anything
fancy that we all can use? Have you already installed some filter for
the tragedy of the IPv6 routing table?

On 28 Jan 2019, at 22:21, Fredy Kuenzler wrote:

> We‘re testing the SLX for a while now but figured that quite a bit
> of the wheel got reinvented, or, to be more precise, the new wheel is
> only halfway round as of today.
>
> Any suggestions for $other_vendor? I actually liked the hardware of
> the SLX - 24*10G and 12*100G.
>
> Meanwhile, BGP4 aggregation is our new buddy... #LifeBeyond750k.
>
> --
> Fredy Kuenzler
> Init7 (Switzerland) Ltd.
> Technoparkstrasse 5
> CH-8406 Winterthur
> Switzerland
>
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

Am 29.01.19 um 10:59 schrieb Jörg Kost:
>> Meanwhile, BGP4 aggregation is our new buddy... #LifeBeyond750k.
>
> Would be interested to know your aggregate / filter criteria,
> anything fancy that we all can use? Have you already installed some
> filter for the tragedy of the IPv6 routing table?
Regarding the aggregation aka #LifeBeyond750k I presented some concept
ideas at SwiNOG 33. See the presentation (May 2018) here.

http://www.swinog.ch/wp-content/uploads/2018/07/Fredy_Kuenzler_life_beyond_750k_init7.pdf

We made further tests since and we are also addressing the IPv6
aggregation with priority these days and should be able to come up with
configuration examples shortly.

--
Fredy Kuenzler

Init7 (Switzerland) Ltd.
AS13030
Technoparkstrasse 5
CH-8406 Winterthur
Twitter: @init7 / @kuenzler
http://www.init7.net/

Fredy, this made me remember the arista spotify stunt: https://labs.spotify.com/2016/01/27/sdn-internet-router-part-2/

Nils

> Am 29.01.19 um 10:59 schrieb Jörg Kost:
> >> Meanwhile, BGP4 aggregation is our new buddy... #LifeBeyond750k.
> >
> > Would be interested to know your aggregate / filter criteria,
> > anything fancy that we all can use? Have you already installed some
> > filter for the tragedy of the IPv6 routing table?
> Regarding the aggregation aka #LifeBeyond750k I presented some concept
> ideas at SwiNOG 33. See the presentation (May 2018) here.
>
> http://www.swinog.ch/wp-
> content/uploads/2018/07/Fredy_Kuenzler_life_beyond_750k_init7.pdf
>
> We made further tests since and we are also addressing the IPv6
> aggregation with priority these days and should be able to come up with
> configuration examples shortly.


_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

Oh well, I am more into blocking certain ASNs and insert a default route
into the network. I took some of the top de-aggregation networks from
the CIDR-reports, that I am not interested in, and generated some
negative list.

Input:
https://github.com/ipcjk/asnbuilder/blob/master/customASN.txt
Output:
https://github.com/ipcjk/asnbuilder/blob/master/saveTheFIB.txt

This saves around 90k IP4 routes and 13k IP6 routes, nothing fancy
there. This can be fine-tuned to bare minimum, enriched with *-flow or
telemetry data and aligned with downstream customers.

So instead of our current positive tracking (which ASN accumulates
what-bytes), we could apply negative tracking (which ASN is not peer, is
not having traffic, ..) and add this to our saveTheFIB as-path. If the
netflow collector sees traffic to a ASN, put the routes back into the
system.

Long-term I am thinking about the SLX9640 as a replacement for our edge
MLXE-4s. Also I often jealously look at the CER-RT-X FIB and cry ;)

Jörg

On 29 Jan 2019, at 11:30, Fredy Kuenzler wrote:

> Am 29.01.19 um 10:59 schrieb Jörg Kost:
>>> Meanwhile, BGP4 aggregation is our new buddy... #LifeBeyond750k.
>>
>> Would be interested to know your aggregate / filter criteria,
>> anything fancy that we all can use? Have you already installed some
>> filter for the tragedy of the IPv6 routing table?
> Regarding the aggregation aka #LifeBeyond750k I presented some concept
> ideas at SwiNOG 33. See the presentation (May 2018) here.
>
> http://www.swinog.ch/wp-content/uploads/2018/07/Fredy_Kuenzler_life_beyond_750k_init7.pdf
>
> We made further tests since and we are also addressing the IPv6
> aggregation with priority these days and should be able to come up
> with
> configuration examples shortly.
>
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

> Long-term I am thinking about the SLX9640 as a replacement for our edge
> MLXE-4s. Also I often jealously look at the CER-RT-X FIB and cry ;)
>
> Jörg

Then you're reminded of the 20 minutes it takes to download the full table?
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

Yes. FIB capability is nice, but CPU is crappy... :-)

Le 15 février 2019 11:30:48 GMT+01:00, Tim Warnock <timoid@timoid.org> a écrit :
>
>> Long-term I am thinking about the SLX9640 as a replacement for our
>edge
>> MLXE-4s. Also I often jealously look at the CER-RT-X FIB and cry ;)
>>
>> Jörg
>
>Then you're reminded of the 20 minutes it takes to download the full
>table?
>_______________________________________________
>foundry-nsp mailing list
>foundry-nsp@puck.nether.net
>http://puck.nether.net/mailman/listinfo/foundry-nsp

--
Envoyé de mon appareil Android avec Courriel K-9 Mail. Veuillez excuser ma brièveté.

The MPC8544 (CER/CES) is actually a successful processor, but maybe the
1 GHz version had not been available during R&D, so they went for the
800 MHz. \_(?)_/¯

The MLX uses the MPC7448 with more beats:
https://www.nxp.com/products/processors-and-microcontrollers/power-architecture-processors/integrated-host-processors/risc-microprocessor:MPC7448

SLX goes all in on Intel Broadwell, so you can run the OS very easily
inside a KVM for toying around.

Happy weekend
J.


On 15 Feb 2019, at 11:36, Clément Cavadore wrote:

> Yes. FIB capability is nice, but CPU is crappy... :-)
>
> Le 15 février 2019 11:30:48 GMT+01:00, Tim Warnock
> <timoid@timoid.org> a écrit :
>>
>>> Long-term I am thinking about the SLX9640 as a replacement for our
>> edge
>>> MLXE-4s. Also I often jealously look at the CER-RT-X FIB and cry ;)
>>>
>>> Jörg
>>
>> Then you're reminded of the 20 minutes it takes to download the full
>> table?
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

Regarding CER-RT
Starting NI 5.8 speed of BGP convergence is much better than in earlier
software releases.
They changed priority of BGP process for scheduler.



On Fri, Feb 15, 2019 at 11:37 AM Clément Cavadore <clement@cavadore.net>
wrote:

> Yes. FIB capability is nice, but CPU is crappy... :-)
>
> Le 15 février 2019 11:30:48 GMT+01:00, Tim Warnock <timoid@timoid.org> a
> écrit :
>>
>>
>> Long-term I am thinking about the SLX9640 as a replacement for our edge
>>> MLXE-4s. Also I often jealously look at the CER-RT-X FIB and cry ;)
>>>
>>> Jörg
>>>
>>
>> Then you're reminded of the 20 minutes it takes to download the full table?
>> ------------------------------
>> foundry-nsp mailing list
>> foundry-nsp@puck.nether.net
>> http://puck.nether.net/mailman/listinfo/foundry-nsp
>>
>>
> --
> Envoyé de mon appareil Android avec Courriel K-9 Mail. Veuillez excuser ma
> brièveté.
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp
>

On Sat, 2019-02-16 at 20:52 +0100, Robert Hass wrote:
> Regarding CER-RT
> Starting NI 5.8 speed of BGP convergence is much better than in
> earlier software releases.
> They changed priority of BGP process for scheduler.

Is it the case for NI 6.0 ? 
I am running that version on CER-RT boxes, and actually, I'd prefer to
have slower convergence (with a default route for example), rather than
a 100% stuck CPU on BGP/IO processes, because it makes other protocols
(OSPF/BGP sessions & so on) flap because of the lack of available CPU
cycles to handle keepalives.
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

Starting several years ago we switched from 1M route cards and chassis
to 2M route cards and chassis, so those sites will be fine for a while
(depending on growth).

That left quite a few XMR 8000s and 4000s out there, which we replaced
with SLX 9540s and 9640s.

We've been not so vocal about it because just like with any brand new
platform we end up being essentially a free outsourced testing service
for our router vendors.

At this point I'm happy with the progress (new software versions with
fixes) enough to post about it.

The 9640s support 4M routes.

Mike.

On 1/28/19 1:31 PM, "Rolf Hanßen" wrote:
> Hi,
>
> I guess most of the users simply replaced their Brocade boxes with
> Juniper, Cisco or Arista and have no need to use that list here anymore.
> ;)
>
> kind regards
> Rolf
>
>> Is there a new list for Foundry/Brocade/Extreme?
>>
>> Thanks,
>>
>> joe
>>
>>
>> Joe McGuckin
>> ViaNet Communications
>>
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

I suppose they could keep some customers if they went with some sort of an
attractive trade up program for X2 cards. There are a lot of networks out
there that don't take a full table, though, and if you study the CIDR
report you will find several networks out there announcing /16 nets and
even larger with a more specific for every single one of their /24 nets
with an identical AS PATH. It looks like the more specifics aren't
multi-homed and they are just worried about route hijacking, I suppose.
But if I filter everything longer than the aggregate, I'll never see a
hijacking attempt on one of the more specifics anyway. One can clobber a
lot of unnecessary routes that way but it's still really a game of whack a
mole. The K series line cards and rack switches from Arista have been
pretty stable so far. Our old MLX units have really been workhorses over
the years, though. We had one failure of an 8x10G card in the past 3 years.

Extreme never really impressed me at layer 3 anyway and I always saw them
as best at layer 2 inside the data center. In fact. the place where I
worked way back in the day was an Extreme shop until I needed a rather
simple BGP option (maximum-paths to do multiple paths to a destination) and
they said BGP really wasn't on their road map and that's when I called
Foundry. It was Extreme that actually pushed me to Foundry and eventually
Brocade in the first place. We're in the process of forklifting the last
of the MLX units out this year, though.


On Fri, Mar 8, 2019 at 4:21 PM Mike Leber <mleber@he.net> wrote:

> Starting several years ago we switched from 1M route cards and chassis
> to 2M route cards and chassis, so those sites will be fine for a while
> (depending on growth).
>
> That left quite a few XMR 8000s and 4000s out there, which we replaced
> with SLX 9540s and 9640s.
>
> We've been not so vocal about it because just like with any brand new
> platform we end up being essentially a free outsourced testing service
> for our router vendors.
>
> At this point I'm happy with the progress (new software versions with
> fixes) enough to post about it.
>
> The 9640s support 4M routes.
>
> Mike.
>
> On 1/28/19 1:31 PM, "Rolf Hanßen" wrote:
> > Hi,
> >
> > I guess most of the users simply replaced their Brocade boxes with
> > Juniper, Cisco or Arista and have no need to use that list here anymore.
> > ;)
> >
> > kind regards
> > Rolf
> >
> >> Is there a new list for Foundry/Brocade/Extreme?
> >>
> >> Thanks,
> >>
> >> joe
> >>
> >>
> >> Joe McGuckin
> >> ViaNet Communications
> >>
> > _______________________________________________
> > foundry-nsp mailing list
> > foundry-nsp@puck.nether.net
> > http://puck.nether.net/mailman/listinfo/foundry-nsp
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp
>

We replaced our Extreme gear with Brocade several years ago. We're
getting ready to swap out some MLXes' with SLXs'.  We've looked at
Arista hard but just can't seem to justify the 3x price premium.

Aaron


On 3/12/2019 3:21 AM, George B wrote:
> I suppose they could keep some customers if they went with some sort
> of an attractive trade up program for X2 cards.  There are a lot of
> networks out there that don't take a full table, though, and if you
> study the CIDR report you will find several networks out there
> announcing /16 nets and even larger with a more specific for every
> single one of their /24 nets with an identical AS PATH.  It looks like
> the more specifics aren't multi-homed and they are just worried about
> route hijacking, I suppose.  But if I filter everything longer than
> the aggregate, I'll never see a  hijacking attempt on one of the more
> specifics anyway. One can clobber a lot of unnecessary routes that way
> but it's still really a game of whack a mole.   The K series line
> cards and rack switches from Arista have been pretty stable so far. 
> Our old MLX units have really been workhorses over the years, though.
> We had one failure of an 8x10G card in the past 3 years.
>
> Extreme never really impressed me at layer 3 anyway and I always saw
> them as best at layer 2 inside the data center.  In fact. the place
> where I worked way back in the day was an Extreme shop until I needed
> a rather simple BGP option (maximum-paths to do multiple paths to a
> destination) and they said BGP really wasn't on their road  map and
> that's when I called Foundry.  It was Extreme that actually pushed me
> to Foundry and eventually Brocade in the first place.  We're in the
> process of forklifting the last of the MLX units out this year, though.
>
>
> On Fri, Mar 8, 2019 at 4:21 PM Mike Leber <mleber@he.net
> <mailto:mleber@he.net>> wrote:
>
> Starting several years ago we switched from 1M route cards and chassis
> to 2M route cards and chassis, so those sites will be fine for a while
> (depending on growth).
>
> That left quite a few XMR 8000s and 4000s out there, which we replaced
> with SLX 9540s and 9640s.
>
> We've been not so vocal about it because just like with any brand new
> platform we end up being essentially a free outsourced testing service
> for our router vendors.
>
> At this point I'm happy with the progress (new software versions with
> fixes) enough to post about it.
>
> The 9640s support 4M routes.
>
> Mike.
>
> On 1/28/19 1:31 PM, "Rolf Hanßen" wrote:
> > Hi,
> >
> > I guess most of the users simply replaced their Brocade boxes with
> > Juniper, Cisco or Arista and have no need to use that list here
> anymore.
> > ;)
> >
> > kind regards
> > Rolf
> >
> >> Is there a new list for Foundry/Brocade/Extreme?
> >>
> >> Thanks,
> >>
> >> joe
> >>
> >>
> >> Joe McGuckin
> >> ViaNet Communications
> >>
> > _______________________________________________
> > foundry-nsp mailing list
> > foundry-nsp@puck.nether.net <mailto:foundry-nsp@puck.nether.net>
> > http://puck.nether.net/mailman/listinfo/foundry-nsp
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net <mailto:foundry-nsp@puck.nether.net>
> http://puck.nether.net/mailman/listinfo/foundry-nsp
>
>
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp

--
================================================================
Aaron Wendel
Chief Technical Officer
Wholesale Internet, Inc. (AS 32097)
(816)550-9030
http://www.wholesaleinternet.com
================================================================

_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

On 3/12/19 1:19 PM, Aaron wrote:
> We replaced our Extreme gear with Brocade several years ago. We're
> getting ready to swap out some MLXes' with SLXs'.  We've looked at
> Arista hard but just can't seem to justify the 3x price premium.

Unless Extreme has massively dropped the price or Arista has massively
increased their prices, there's no reason that equivalent Arista gear
should be 3x the price of an SLX, and that doesn't match up with my
experience.

I ended up with the Arista 7280SR (48-ports of 10G, 6 ports of
short-range 100G in a 1RU fixed configuration) and it's been working well.

--
Richard
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

I just got quotes for an Arista 7280R (48x 10Gbit and 6x 100Gbit) and an
SLX9640.  Adding a port license to the 9640 to go to 8x 100G ports the
Arista quote for just the base box (no service, optics, etc.) was 251%
higher.

Maybe I have the wrong sales rep at Arista and the right one at Extreme?

Aaron


On 3/12/2019 1:39 PM, Richard Laager wrote:
> On 3/12/19 1:19 PM, Aaron wrote:
>> We replaced our Extreme gear with Brocade several years ago. We're
>> getting ready to swap out some MLXes' with SLXs'.  We've looked at
>> Arista hard but just can't seem to justify the 3x price premium.
> Unless Extreme has massively dropped the price or Arista has massively
> increased their prices, there's no reason that equivalent Arista gear
> should be 3x the price of an SLX, and that doesn't match up with my
> experience.
>
> I ended up with the Arista 7280SR (48-ports of 10G, 6 ports of
> short-range 100G in a 1RU fixed configuration) and it's been working well.
>

--
================================================================
Aaron Wendel
Chief Technical Officer
Wholesale Internet, Inc. (AS 32097)
(816)550-9030
http://www.wholesaleinternet.com
================================================================

_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

Rack space is the premium with me. A 7280R2K-48C6 gives me 48x10G plus
6x40/100G ports in 1U with 2 million routes vs 7U for an MLX-8.


On Tue, Mar 12, 2019, 11:49 Aaron <aaron@wholesaleinternet.net> wrote:

> I just got quotes for an Arista 7280R (48x 10Gbit and 6x 100Gbit) and an
> SLX9640. Adding a port license to the 9640 to go to 8x 100G ports the
> Arista quote for just the base box (no service, optics, etc.) was 251%
> higher.
>
> Maybe I have the wrong sales rep at Arista and the right one at Extreme?
>
> Aaron
>
>
> On 3/12/2019 1:39 PM, Richard Laager wrote:
> > On 3/12/19 1:19 PM, Aaron wrote:
> >> We replaced our Extreme gear with Brocade several years ago. We're
> >> getting ready to swap out some MLXes' with SLXs'. We've looked at
> >> Arista hard but just can't seem to justify the 3x price premium.
> > Unless Extreme has massively dropped the price or Arista has massively
> > increased their prices, there's no reason that equivalent Arista gear
> > should be 3x the price of an SLX, and that doesn't match up with my
> > experience.
> >
> > I ended up with the Arista 7280SR (48-ports of 10G, 6 ports of
> > short-range 100G in a 1RU fixed configuration) and it's been working
> well.
> >
>
> --
> ================================================================
> Aaron Wendel
> Chief Technical Officer
> Wholesale Internet, Inc. (AS 32097)
> (816)550-9030
> http://www.wholesaleinternet.com
> ================================================================
>
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp
>

The SLX9640 is a 1U box with 24x 10G ports and 12x 100G ports (4
licensed on the base config) with 4 million routes for less than a third
of the price.

I'd love to try Arista but that's what I'm dealing with.

Aaron


On 3/12/2019 5:46 PM, George B wrote:
> Rack space is the premium with me. A 7280R2K-48C6 gives me 48x10G plus
> 6x40/100G ports in 1U with 2 million routes vs 7U for an MLX-8.
>
>
> On Tue, Mar 12, 2019, 11:49 Aaron <aaron@wholesaleinternet.net
> <mailto:aaron@wholesaleinternet.net>> wrote:
>
> I just got quotes for an Arista 7280R (48x 10Gbit and 6x 100Gbit)
> and an
> SLX9640.  Adding a port license to the 9640 to go to 8x 100G ports
> the
> Arista quote for just the base box (no service, optics, etc.) was
> 251%
> higher.
>
> Maybe I have the wrong sales rep at Arista and the right one at
> Extreme?
>
> Aaron
>
>
> On 3/12/2019 1:39 PM, Richard Laager wrote:
> > On 3/12/19 1:19 PM, Aaron wrote:
> >> We replaced our Extreme gear with Brocade several years ago. We're
> >> getting ready to swap out some MLXes' with SLXs'. We've looked at
> >> Arista hard but just can't seem to justify the 3x price premium.
> > Unless Extreme has massively dropped the price or Arista has
> massively
> > increased their prices, there's no reason that equivalent Arista
> gear
> > should be 3x the price of an SLX, and that doesn't match up with my
> > experience.
> >
> > I ended up with the Arista 7280SR (48-ports of 10G, 6 ports of
> > short-range 100G in a 1RU fixed configuration) and it's been
> working well.
> >
>
> --
> ================================================================
> Aaron Wendel
> Chief Technical Officer
> Wholesale Internet, Inc. (AS 32097)
> (816)550-9030
> http://www.wholesaleinternet.com
> ================================================================
>
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net <mailto:foundry-nsp@puck.nether.net>
> http://puck.nether.net/mailman/listinfo/foundry-nsp
>
>
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp

--
================================================================
Aaron Wendel
Chief Technical Officer
Wholesale Internet, Inc. (AS 32097)
(816)550-9030
http://www.wholesaleinternet.com
================================================================

_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

That's what I get for not keeping up with that product line.

I think price difference might also vary according to what kind of
discounts people's vendors can get them. Looking around I don't see that
much of a price difference between the two units though the SLX is lower
priced. About the only real technical advantage I see with the Aristas are
that they have generally been rather free of bugs and the mLAG seems to be
a little better suited to our environment than the MCT of the Brocade
line. That said, I wouldn't turn down a chance to try out one of those SLX
units.

On Tue, Mar 12, 2019 at 3:57 PM Aaron <aaron@wholesaleinternet.net> wrote:

> The SLX9640 is a 1U box with 24x 10G ports and 12x 100G ports (4
> licensed on the base config) with 4 million routes for less than a third
> of the price.
>
> I'd love to try Arista but that's what I'm dealing with.
>
> Aaron
>
>
> On 3/12/2019 5:46 PM, George B wrote:
> > Rack space is the premium with me. A 7280R2K-48C6 gives me 48x10G plus
> > 6x40/100G ports in 1U with 2 million routes vs 7U for an MLX-8.
> >
> >
> > On Tue, Mar 12, 2019, 11:49 Aaron <aaron@wholesaleinternet.net
> > <mailto:aaron@wholesaleinternet.net>> wrote:
> >
> > I just got quotes for an Arista 7280R (48x 10Gbit and 6x 100Gbit)
> > and an
> > SLX9640. Adding a port license to the 9640 to go to 8x 100G ports
> > the
> > Arista quote for just the base box (no service, optics, etc.) was
> > 251%
> > higher.
> >
> > Maybe I have the wrong sales rep at Arista and the right one at
> > Extreme?
> >
> > Aaron
> >
> >
> > On 3/12/2019 1:39 PM, Richard Laager wrote:
> > > On 3/12/19 1:19 PM, Aaron wrote:
> > >> We replaced our Extreme gear with Brocade several years ago. We're
> > >> getting ready to swap out some MLXes' with SLXs'. We've looked at
> > >> Arista hard but just can't seem to justify the 3x price premium.
> > > Unless Extreme has massively dropped the price or Arista has
> > massively
> > > increased their prices, there's no reason that equivalent Arista
> > gear
> > > should be 3x the price of an SLX, and that doesn't match up with my
> > > experience.
> > >
> > > I ended up with the Arista 7280SR (48-ports of 10G, 6 ports of
> > > short-range 100G in a 1RU fixed configuration) and it's been
> > working well.
> > >
> >
> > --
> > ================================================================
> > Aaron Wendel
> > Chief Technical Officer
> > Wholesale Internet, Inc. (AS 32097)
> > (816)550-9030
> > http://www.wholesaleinternet.com
> > ================================================================
> >
> > _______________________________________________
> > foundry-nsp mailing list
> > foundry-nsp@puck.nether.net <mailto:foundry-nsp@puck.nether.net>
> > http://puck.nether.net/mailman/listinfo/foundry-nsp
> >
> >
> > _______________________________________________
> > foundry-nsp mailing list
> > foundry-nsp@puck.nether.net
> > http://puck.nether.net/mailman/listinfo/foundry-nsp
>
> --
> ================================================================
> Aaron Wendel
> Chief Technical Officer
> Wholesale Internet, Inc. (AS 32097)
> (816)550-9030
> http://www.wholesaleinternet.com
> ================================================================
>
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp
>

To other platforms?

We still have a few of these running, and we appear to be about to hit the
wall.

What's going on here?

System Parameters Default Maximum Current Actual Bootup Revertible
ip-cache 204800 1048576 1048576 1048576 1048576 Yes
ip-route 204800 1048576 1048576 1048576 1048576 Yes
ipv6-cache 65536 245760 245760 245760 245760 Yes
ipv6-route 65536 245760 245760 245760 245760 Yes

Router has been rebooted since those system-max limits were changed.

#show cam-partition slot 1

CAM partitioning profile: ipv4-ipv6-2

Slot 1 XPP20SP 0:
# of CAM device = 4
Total CAM Size = 917504 entries (63Mbits)

IP: Raw Size 786432, User Size 786432(0 reserved)
Subpartition 0: Raw Size 2420, User Size 2420, (0 reserved)
Subpartition 1: Raw Size 692807, User Size 692807, (0 reserved)
Subpartition 2: Raw Size 73329, User Size 73329, (0 reserved)
Subpartition 3: Raw Size 13567, User Size 13567, (0 reserved)
Subpartition 4: Raw Size 3075, User Size 3075, (0 reserved)

IPv6: Raw Size 131072, User Size 65536(0 reserved)
Subpartition 0: Raw Size 188, User Size 94, (0 reserved)
Subpartition 1: Raw Size 121688, User Size 60844, (0 reserved)
Subpartition 2: Raw Size 7940, User Size 3970, (0 reserved)
Subpartition 3: Raw Size 1058, User Size 529, (0 reserved)
Subpartition 4: Raw Size 160, User Size 80, (0 reserved)

If I'm interpreting that output correctly, the router has a total of
917504 CAM entries for routes, it's split between v4/v6, and in spite of
the system-max settings, this router will be out of FIB space at 786432 v4
routes / 65536 v6 routes...or any moment now for v6. This router is
logging CAM partition warnings about free slots being low.

We're seeing the same on MLXe and NetIron chassis in XMR mode
running 5.6.0j.


----------------------------------------------------------------------
Jon Lewis, MCP :) | I route
| therefore you are
_________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

On 13 Mar 2019, at 16:03, Jon Lewis wrote:
> If I'm interpreting that output correctly, the router has a total of
> 917504 CAM entries for routes, it's split between v4/v6, and in spite
> of the system-max settings, this router will be out of FIB space at
> 786432 v4 routes / 65536 v6 routes...or any moment now for v6. This
> router is logging CAM partition warnings about free slots being low.

Something around this, I think it was even 768000 v4 routes only.

Can you check the output of:

show ip cache
show ipv6 cache
show ip route summary
show ipv6 route summary


_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

The XMR and MLXe are limited to 786432 IPv4 routes and 65536 for IPv6 (ipv4-ipv6-2 profile)
If you want more routes you need -X2 linecards.

The IPv4 table and IPv6 tables are close to limit from the XMR/MLXe -x cards (+-757166 ipv4 and +-65239 ipv6)
That is why you get the cam profile warnings.



Kind regards/Met vriendelijke groet,

Dennis op de Weegh

?

Bitency
Hoge Ham 24
5104JG Dongen
?
Kvk nummer: 20144338
BTW nummer: NL213538519B01
?
W: www.bitency.nl
E: info@bitency.nl
T: +31 (0)162 714066


-----Oorspronkelijk bericht-----
Van: foundry-nsp <foundry-nsp-bounces@puck.nether.net> Namens Jon Lewis
Verzonden: woensdag 13 maart 2019 16:04
Aan: foundry-nsp@puck.nether.net
Onderwerp: Re: [f-nsp] Where have all the MLX/XMR users gone to?

To other platforms?

We still have a few of these running, and we appear to be about to hit the
wall.

What's going on here?

System Parameters Default Maximum Current Actual Bootup Revertible
ip-cache 204800 1048576 1048576 1048576 1048576 Yes
ip-route 204800 1048576 1048576 1048576 1048576 Yes
ipv6-cache 65536 245760 245760 245760 245760 Yes
ipv6-route 65536 245760 245760 245760 245760 Yes

Router has been rebooted since those system-max limits were changed.

#show cam-partition slot 1

CAM partitioning profile: ipv4-ipv6-2

Slot 1 XPP20SP 0:
# of CAM device = 4
Total CAM Size = 917504 entries (63Mbits)

IP: Raw Size 786432, User Size 786432(0 reserved)
Subpartition 0: Raw Size 2420, User Size 2420, (0 reserved)
Subpartition 1: Raw Size 692807, User Size 692807, (0 reserved)
Subpartition 2: Raw Size 73329, User Size 73329, (0 reserved)
Subpartition 3: Raw Size 13567, User Size 13567, (0 reserved)
Subpartition 4: Raw Size 3075, User Size 3075, (0 reserved)

IPv6: Raw Size 131072, User Size 65536(0 reserved)
Subpartition 0: Raw Size 188, User Size 94, (0 reserved)
Subpartition 1: Raw Size 121688, User Size 60844, (0 reserved)
Subpartition 2: Raw Size 7940, User Size 3970, (0 reserved)
Subpartition 3: Raw Size 1058, User Size 529, (0 reserved)
Subpartition 4: Raw Size 160, User Size 80, (0 reserved)

If I'm interpreting that output correctly, the router has a total of
917504 CAM entries for routes, it's split between v4/v6, and in spite of
the system-max settings, this router will be out of FIB space at 786432 v4
routes / 65536 v6 routes...or any moment now for v6. This router is
logging CAM partition warnings about free slots being low.

We're seeing the same on MLXe and NetIron chassis in XMR mode
running 5.6.0j.


----------------------------------------------------------------------
Jon Lewis, MCP :) | I route
| therefore you are
_________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

On Fri, 15 Mar 2019, Dennis op de Weegh wrote:

> The XMR and MLXe are limited to 786432 IPv4 routes and 65536 for IPv6 (ipv4-ipv6-2 profile)
> If you want more routes you need -X2 linecards.
>
> The IPv4 table and IPv6 tables are close to limit from the XMR/MLXe -x cards (+-757166 ipv4 and +-65239 ipv6)
> That is why you get the cam profile warnings.

:(

I'm going to lab test it, but I was hoping that more sensible system-max
settings (keeping in mind there's only 1M slots to carve up) would make it
possible to raise the v6 limit to around 75k while still maintaining
enough room for some v4 table growth, the idea being, buy some time so we
don't have to rush to decom our remaining brocade routers or come up with
other workarounds (less than full tables).

----------------------------------------------------------------------
Jon Lewis, MCP :) | I route
| therefore you are
_________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

On Fri, 15 Mar 2019, J?rg Kost wrote:

> On 13 Mar 2019, at 16:03, Jon Lewis wrote:
>> If I'm interpreting that output correctly, the router has a total of
>> 917504 CAM entries for routes, it's split between v4/v6, and in spite of
>> the system-max settings, this router will be out of FIB space at 786432 v4
>> routes / 65536 v6 routes...or any moment now for v6. This router is
>> logging CAM partition warnings about free slots being low.
>
> Something around this, I think it was even 768000 v4 routes only.
>
> Can you check the output of:
>
> show ip cache
> show ipv6 cache
> show ip route summary
> show ipv6 route summary

I'm not where I can easily get at all that info, but I do have the first
two available.

Total IP and IPVPN Cache Entry Usage on LPs:
Module Host Network Free Total
1 826 755362 292388 1048576
2 2181 755362 291033 1048576

Total IPv6 and IPv6 VPN Cache Entry Usage on LPs:
Module Host Network Free Total
1 151 66067 251738 317956
2 150 66067 251739 317956

These numbers, esp the IPv6, don't really make sense, since the v6 ones
imply the router has reserved more space for IPv6 route-cache than the
system-max settings requested...and regardless of what the show ip[v6]
cache output says, the router seems to be out of slots. It's got a couple
of full transit feeds and a couple dozen public peers...so roughly 755k v4
routes and roughly 66k v6 routes.

----------------------------------------------------------------------
Jon Lewis, MCP :) | I route
| therefore you are
_________ http://www.lewis.org/~jlewis/pgp for PGP public key_________

On Fri, 15 Mar 2019, Dennis op de Weegh wrote:

> The XMR and MLXe are limited to 786432 IPv4 routes and 65536 for IPv6 (ipv4-ipv6-2 profile)
> If you want more routes you need -X2 linecards.
>
> The IPv4 table and IPv6 tables are close to limit from the XMR/MLXe -x cards (+-757166 ipv4 and +-65239 ipv6)
> That is why you get the cam profile warnings.

It seems Brocade was even worse than Cisco (1M routes on the 3bxl), in
that their 1M route cards are really 1M routes - 2x v6 routes - about
100000 TCAM slots for other misc uses, and you can't fine tune the v4/v6
split. I finally got a chance to do some messing around with a lab XMR
with full BGP feeds, and was I think I found is:

system-max settings dealing with max numbers of ip[v6]-route/cache affect
what the router is willing/able to store in RIB, but the cam-partition
profile affects what can be stuffed in FIB, and other than selecting
between the various cam-partition profiles, the TCAM carving between v4/v6
is not flexible (or if it is, that's a hidden command). So, for a
full-table router, the only cam-partition profile of any use is
ipv4-ipv6-2, and that profile may or may not hold a "full" table today
depending on the exact size of your full table.

Or in other words, the MLX/XMR is dead or very nearly there for anyone
needing full v4 & v6 tables.

----------------------------------------------------------------------
Jon Lewis, MCP :) | I route
| therefore you are
_________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

XMR life time has or will come very soon for DFZ routing, the MLXe (if
you have one) is still a valid platform for DFZ and I am sure it is
going to be supported at least till 2025/2026.

So configurations that still work:
- MLXE chassis with MR2-X-management cards

Modules:
- BR-MLX-100Gx2-CFP2-X2
-> 2 x 100g
- BR-MLX-10GX20-X2 in its variants:
-> BR-MLX-10Gx20-X2
-> 20 x 1g/10g
-> BR-MLX-10Gx10-X2
-> 10 x 1g/10g, upgradeable to 20 1g/20g
-> BR-MLX-1GX10-U10G-X2
-> 20 x 1g, upgradeable to 20 1g/20g

Cam profiles:
- default (1424K mil ipv4, 416K ipv6)
- multiservice-6 ( 1120K ipv4, 768K ipv6)
- ipv4-ipv6-2 (2048K ipv4, 1024K ipv6, no VPN)
-> reference
https://documentation.extremenetworks.com/netiron/SW/63x/6300a/netIron-6300a-managementguide.pdf

See also
https://gtacknowledge.extremenetworks.com/articles/Q_A/How-to-enable-2-Million-routes-on-MLX-platform

If you buy new, SLX 9540/9640/9850 are valid targets but this leads the
thread back to the beginning ;-).


On 22 Mar 2019, at 0:37, Jon Lewis wrote:

> On Fri, 15 Mar 2019, Dennis op de Weegh wrote:
>
>> The XMR and MLXe are limited to 786432 IPv4 routes and 65536 for IPv6
>> (ipv4-ipv6-2 profile)
>> If you want more routes you need -X2 linecards.
>>
>> The IPv4 table and IPv6 tables are close to limit from the XMR/MLXe
>> -x cards (+-757166 ipv4 and +-65239 ipv6)
>> That is why you get the cam profile warnings.
>
> It seems Brocade was even worse than Cisco (1M routes on the 3bxl), in
> that their 1M route cards are really 1M routes - 2x v6 routes - about
> 100000 TCAM slots for other misc uses, and you can't fine tune the
> v4/v6 split. I finally got a chance to do some messing around with a
> lab XMR with full BGP feeds, and was I think I found is:
>
> system-max settings dealing with max numbers of ip[v6]-route/cache
> affect what the router is willing/able to store in RIB, but the
> cam-partition profile affects what can be stuffed in FIB, and other
> than selecting between the various cam-partition profiles, the TCAM
> carving between v4/v6 is not flexible (or if it is, that's a hidden
> command). So, for a full-table router, the only cam-partition profile
> of any use is ipv4-ipv6-2, and that profile may or may not hold a
> "full" table today depending on the exact size of your full table.
>
> Or in other words, the MLX/XMR is dead or very nearly there for anyone
> needing full v4 & v6 tables.
>
> ----------------------------------------------------------------------
> Jon Lewis, MCP :) | I route
> | therefore you are
> _________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp