[c-nsp] catalyst 3560 performance issue

[Tassos Chatzithomaoglou]

There is always the case that some traffic needs to be software swithed, so you might have a problem 
then.
I have a 3550 (doing both L2/L3) which gets >80% cpu usage when the traffic reaches ~70Mbps (70 in, 
70 out). Still trying to figure out what kind of traffic causes this increase in cpu load.

--
Tassos

Marko Milivojevic wrote on 29/6/2006 2:05 μμ:
> You will experience no performance degradation with ~100 Mb/s. We have 
> several 3560 routing and switching much more than that.
> 
> As Asbjorn already mentioned, SDM template would affect performance only 
> when resources (memory) have been exhausted, which is most probably not the 
> case. How many routes/MAC addresses will you have?
> 
> 
> Marko.
> 
> Vladimir Rabljenovic wrote:
>> Hello,
>>  
>> Thank you for the answer. I will try to be more precise. 
>> I would like to know how this relate to situation when switch acts also as L3 device (in parallel to
>> L2 switching). 
>> I have a situation that around 100Mbps of traffic is expected to come via 2Gig port-channel trunk,
>> then L2 switched to a few FastEth interfaces. After that, this traffic is coming back on some other
>> fasteth interfaces into switch, and is L3 routed towards mentioned 2Gig port-channel via different
>> Vlan. 
>> So, when using desktop routing sdm template, I am interested whether I could experience some
>> performance problems or not. 
>> Unforunately, I am not currently in the position to run some extensive test with traffic generators.
>> If someone has some experience on this, I would very much appreciate to share it with me.
>>
>> thanks in advance
>> vlado
>>
>>
>>>> In general, I am interested about info about performance
>>>> comparison when different sdm templates/scenarios are deployed.
>>> The choice of SDM template doesn't (directly) affect the
>>> forwarding performance, only how the (obviously finite) amount
>>> of hardware resources are sliced up for different purposes.
>>>
>>> Of cause, if you slice it up incorrectly, it may adversely impact
>>> performance when you run out of some ressource.
>>>
>>> See http://tinyurl.com/e7k8w
>>>
>>> -A
>>>
>> </blockquote>
>>
>>
> _______________________________________________
> 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/