[c-nsp] Nightmare for load balancing of L2VPN traffic on CRS (traffic from ME3600)

[Tassos Chatzithomaoglou]

We have been fighting with Cisco for this (L2VPN PW load balancing) for a long time.
Besides a good (which could have been much better) level of support on the ASR9k, it's a shame that all other new equipment doesn't even support FAT or any other good alternative.
PS: As a temporary workaround, we have increased the number of PWs accross the network.

--
Tassos

Darren Liew wrote on 13/4/2015 4:33 πμ:
> Hi Adam & Mark,
>
> We have opened Cisco TAC case to check, apparently the FAT-PW feature is
> not on Cisco ME3600 roadmap for now.
>
> Even with FAT-PW solution, it doesn't sounds ideal as it requires other
> equipment to change config to overcome the shortcoming of CRS. Imagine we
> already have few thousands of PW around the network.
>
> We are talking to Cisco on this. Will keep you guys posted of the outcome.
> Cheers !
>
> Rgds
> Darren Liu
>
>
>
> On Sun, Apr 12, 2015 at 10:29 PM, Mark Tinka <mark.tinka at seacom.mu> wrote:
>
>>
>> On 9/Apr/15 09:31, Adam Vitkovsky wrote:
>>> Hi Darren,
>>>
>>> Been in the same shituation though I'm not familiar with CRS MPLS
>> hashing algo I guess it also looks at the next header field after the
>> bottom of the stack label and if it won't find 0x4 or 0x6 (hopefully in
>> case of L2VPNs) it won't do src/dst IP address hash so for L2VPNs the
>> hashing is done solely based on the VC label which is a bummer.
>>> And yes I've been asking for FAT-PW on MEs some time ago and there's no
>> will at all...
>>> In our case we carried a bulk of VLANs in the high bandwidth PWs so we
>> broke them into ever smaller PWs and that kind of helped to balance the
>> traffic a little better.
>>> Otherwise without FAT-PW I guess the only option is to get MPLS-TE
>> involved then you can tell which PW should use which TE tunnel thus balance
>> the traffic "evenly" across the backbone.
>>
>> This thread is timely, as I was looking at FAT (RFC 6391) just last week.
>>
>> So implementation in IOS and IOS XE seems cumbersome, i.e., there is
>> some kind of "interface pseudowire" thing that looks clumsy, and yet the
>> IOS XR implementation is more as one would expect.
>>
>> At any rate, I see the code in the ME3600X software, but have no tried
>> it. Are folk saying the code is there but it doesn't execute?
>>
>> Mark.
>>
> _______________________________________________
> 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/
>