[c-nsp] IP multicast traffic overwhelms switches

victor vitya at list.ru
Fri Jul 10 15:35:17 EDT 2009


On Fri, 10 Jul 2009 23:03:08 +0400, Jay Ford <jay-ford at uiowa.edu> wrote:

> In your original message you said:
>     I explicitly entered "ip mroute-cache" under every interface
> which I took to mean that you were changing the default.
>
> In my experience on 6500 boxes running various 12.2SX versions "ip
> mroute-cache" does not show up by default.
>
Right. Let me explain. After I configured initial setup I ran a test that  
showed intensive cpu usage. My first thought was about process switching  
and then I entered "ip mroute-cache" for the interfaces. But the command  
do not appear when you do "sho run int" that made me think it is on by  
default.
> If you do "show ip interface" for your edge-facing interfaces, does
>     IP multicast multilayer switching is enabled
> appear near the end of the output?
>
No, couldn't find it
> Also, does "show mls ip multicast" show your multicast traffic being  
> hardware switched?

#sho mls ip m
Multicast hardware switched flows:
Total hardware switched flows : 0

>
>>> Are the 4900 boxes doing L3 or just L2?  I suspect they'd do much  
>>> better at
>>> L2 fan-out of multicast than at L3 fan-out.  You're probably hitting a  
>>> pps
>>> or
>>> packet replication limit before hitting the bps limit.
>> I agree that this switch will probably perform better doing L2 exchange  
>> but
>> then there is another problem: C7604 carry QinQ vlans and C4924  
>> terminates
>> them giving each tunnel's payload out of a deferent dot1q-tunnel access  
>> port.
>> If I don't do multicast routing I will need to carry the same multicast
>> traffic on every configured outer vlan. This will eat up all the  
>> bandwidth.
>
> Bummer, dude.  I don't have anything to offer about that, other than to
> speculate that the QinQ tunnel stuff might be undermining the ability of  
> 1 or
> both boxes to efficiently deal with multicast traffic.  You might have to
> get your Cisco support people in on this one.
>



-- 
Using Opera's revolutionary e-mail client: http://www.opera.com/mail/


More information about the cisco-nsp mailing list