[c-nsp] VSS - Horror stories, show-stoppers, other personal experience?

[Bradley Williamson]

My understanding of an LTL is a path map for traffic through the switch.

In the VSS environment, every multicast stream consumes 3 multicast LTL
resources for each VLAN on the switch and another 3 LTL's for each IGMP
snooping entry. We had 300+ multicast channels coming in from encoders,
then being joined by encryption servers and ICC/error correcting servers.
The encryption servers were sending an encrypted multicast stream back
into the VSS. We also had another set of servers joining the encrypted
streams and generating a PIP stream.

In a stand alone 6509 there are ~30K multicast LTL's available. In the VSS
this number drops to ~20 for the two combined 6509's. Our 300+ channels
with all of the extra stream for PIP, encryption and ICC etc.. Were using
over 19,800 multicast LTL's.

There is an LTL threshold of 200 free LTL's. Once you have less than 200
free multicast LTL's available, the switch will disable IGMP snooping and
begin to flood multicast out of all ports. This obviously crashes the
switch processor. 

Some commands to see LTL usage:

"show platform hardware capacity multicast"
"remote command switch show mmls mltl"

I hope this information helps.

Regards, 

Brad

On 6/17/11 2:50 PM, "Youssef El Fathi" <youssef.el.fathi at gmail.com> wrote:

>Hello Bradley,
>
>I would like to have more details about the performance issue you ran with
>IP mcast traffic? Can you give me more details.
>
>Thanks and regards
>
>Youssef
>
>
>> Message: 8
>> Date: Fri, 17 Jun 2011 08:15:28 -0500
>> From: Bradley Williamson <bwilliamson at eatel.com>
>> To: Mike G <geezyx at gmail.com>
>> Cc: "cisco-nsp at puck.nether.net" <cisco-nsp at puck.nether.net>
>> Subject: Re: [c-nsp] VSS - Horror stories, show-stoppers, other
>>        personal experience?
>> Message-ID: <BB407316-BE74-4E7F-856B-8CB6B53BDC5D at eatel.com>
>> Content-Type: text/plain; charset="us-ascii"
>>
>> I just spent the better part of my day splitting a vss.
>>
>> I think it works well for the most part. Fail over works well. It Is
>>easy
>> to manage. MEC is nice too.
>>
>> We tried it in an Multicast environment, and it was too resource limited
>> for what we were doing. If you are not doing much multicast (300+
>>channels)
>> then it should work well for you.
>>
>> Sent from my iPad
>>
>>
>_______________________________________________
>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/
>