[c-nsp] Recovery time under interface failure - VPLS - MPLS L3 VPN- Plenty L3

[Oliver Boehmer (oboehmer)]

alaerte.vidali at nsn.com <mailto:alaerte.vidali at nsn.com> wrote on Monday,
June 30, 2008 4:24 PM:

> Tks Oliver,
> 
>>> assuming there is no STP delay (portfast/etc.) this should be
>>> rather quick 
> 
> That is as I see it should work for VPLS. But crazily, it is taking 19
> to 20 seconds, even though portfast is enabled.
> 
> Any clue?

No. have you traced in which direction the delay is occuring? I.e. how
long does it take until the client's packet gets to the server (whose
mac-addr should already be known)?

	oli


> -----Original Message-----
> From: ext Oliver Boehmer (oboehmer) [mailto:oboehmer at cisco.com]
> Sent: Monday, June 30, 2008 10:23 AM
> To: Vidali Alaerte (NSN - BR/Rio de Janeiro);
> cisco-nsp at puck.nether.net 
> Subject: RE: [c-nsp] Recovery time under interface failure - VPLS -
> MPLS 
> L3 VPN- Plenty L3
> 
> alaerte.vidali at nsn.com <mailto:alaerte.vidali at nsn.com> wrote on
> Monday, 
> June 30, 2008 2:55 PM:
> 
>> Hi Oliver,
>> 
>> Question is very specific to issue in ethernet connection between R1
>> and laptop and comparing recovery time under this failure in VPLS,
>> L3MPLS VPN and pure L3 routing. That is, how VPLS will influence the
>> recovery of the MAC address on R1 
> 
>> (delay introduced by VPLS).
> 
> Ah, so you mean how long it takes when the ethernet comes back up and
> the client can resume connecting to the server? Well, assuming there
> is 
> no STP delay (portfast/etc.) this should be rather quick (similar to a
> switched environment, with a bit higher propagation delay if the VPLS
> spans a large geography). It's the same flooding/learning method.
> 
> When it comes to routing (L3VPN or "regular"), I guess it depends on
> wether the route to the destination is already known or not. If not,
> getting routing information over to the other side can take ~5-8
> seconds 
> for regular (untuned) IGP or even more for L3VPN (depends on BGP MRAI
> timers and RD-setup/import-delay).
> 
> 	oli
> 
> 
> 
>> 
>> -----Original Message-----
>> From: ext Oliver Boehmer (oboehmer) [mailto:oboehmer at cisco.com]
>> Sent: Sunday, June 29, 2008 4:48 AM
>> To: Vidali Alaerte (NSN - BR/Rio de Janeiro);
>> cisco-nsp at puck.nether.net
>> Subject: RE: [c-nsp] Recovery time under interface failure - VPLS -
>> MPLS L3 VPN- Plenty L3
>> 
>> alaerte.vidali at nsn.com <> wrote on Saturday, June 28, 2008 5:21 PM:
>> 
>>>  Hi,
>>> 
>>> Considering following simple topology:
>>> 
>>> Laptop-----(e1)R1-----R2----R3----Server
>>> 
>>> ...and that OSPF timers are the same and BFD is not used (no failure
>>> recovery optimization used) on all scenarios:
>>> 
>>> What would be the recovery time when interface etherne 1 (from
>>> laptop 
> 
>>> to R1) fails in these cases:
>>> 
>>> -Just IP routing between R1 and R3
>>> -VPLS between R1 and R3
>>> -MPLS VPN between R1 and R3
>>> 
>>> If I am not wrong, in VPLS case R1 will remove MAC address and
>>> communicate that to R3, but not sure if it will impact final
>>> connectivity recovery time between laptop--server.
>>> (sorry, no lab to test right now)
>> 
>> Hmm, where should above topology recover to? There is no alternate
>> path between client and server here? In general, convergence times
>> depend on several variables.. sub-10-sec is a ballpark figure you can
>> use, but it can also take longer (when BGP is involved) or quicker..
>> 
>> 	oli