[c-nsp] SPAN destination on oversubscribed module

[Sigurbjörn Birkir Lárusson]

I was aware that the FPGA pairs were 1,4, 2,3, 5,7 and 6,8 I just reversed
the numbers 3 and 4, sorry about that

Kind regards,
Sibbi

On 9.1.2014 19:33, "Jiri Prochazka" <jiri.prochazka at superhosting.cz> wrote:

>Hi,
>
>mapping of ports on WS-X6708-10GE to ASIC's is quite different. It's
>very common mistake.
>
>First ASIC is mapped to ports ((1,4),(5,7)), second ASIC is mapped to
>ports ((2,3),(6,8)).
>
>Capacity of inner pairs is 16Gbit, quadruples are 20Gbit.
>
>
>You can try to switch module to the performance mode (oversubscribed
>ports will be shutdowned - "no hw-module slot x oversubscription"
>
>It will disable ports 4,7,3,8.
>
>
>All the best,
>
>
>
>Jiri
>
>
>
>Dne 9.1.2014 16:22, Sigurbjörn Birkir Lárusson napsal(a):
>> It¹s a given the load coming from other modules would go through the
>> fabric, whether inter module traffic does depends on which ports you are
>> using and how SPAN is handled with distributed-forwarding, must admit
>>I¹ve
>> never given that much thought.
>>
>> On the 6708 there are two ASICS ,one for port 1,3,5,7 and the other for
>> 2,4,6,8 as I understood it.  Traffic traveling between the ASICs needs
>>to
>> go through the Fabric, Fabric within the ASIC does not.
>>
>> I¹m assuming the situation on the 6716 is similar except then probably
>> with 4 asics instead of 2, so if the mirroring port and the monitoring
>> ports are all on the same asic, that bandwidth might be freed from going
>> over the fabric.  I¹d try this in a lab before committing it though,
>>these
>> cards are a bit dangerous.  Should and does are not the same thing with
>> the 7600 platform (ever).
>>
>> Kind regards,
>> Sibbi
>>
>> On 9.1.2014 15:10, "Ben Hammadi, Kayssar (NSN - TN/Tunis)"
>> <kayssar.ben_hammadi at nsn.com> wrote:
>>
>>> Thanks Phil,
>>>
>>>    The last question is about the effect of the SPAN on fabric load ,
>>>the
>>> monitored traffic from other module will cross the fabric on its way to
>>> the replication engine of module WS-X6716-10T  and contribute on fabric
>>> load right ?
>>>
>>> Br.
>>>
>>> BEN HAMMADI Kayssar
>>>
>>> NOKIA SOLUTION NETWORKS
>>> Lead Engineer -BroadBand Connectivity
>>> JNCIE-M (#471), JNCIE-SP (#1147), CCIP
>>> Mobile : +216 29 349 952  /  +216 98 349 952
>>> Skype : kayssar ben hammadi
>>>
>>>
>>> -----Original Message-----
>>> From: cisco-nsp [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of
>>> ext Phil Mayers
>>> Sent: Thursday, January 09, 2014 3:57 PM
>>> To: cisco-nsp at puck.nether.net
>>> Subject: Re: [c-nsp] SPAN destination on oversubscribed module
>>>
>>> On 09/01/14 14:47, Ben Hammadi, Kayssar (NSN - TN/Tunis) wrote:
>>>> Dears ,
>>>>
>>>> We have two 6509 interconnected via port-channel of two 10Ge port
>>>> carrying about 3Gb. the module used is WS-X6716-10T and only the two
>>>> mentioned port are used on this module for the moment . We are now
>>>> adding a monitoring probe that will be connected on this module , the
>>>> estimated monitored traffic is about 8Gb . Since this module has an
>>>> oversubscription rate of 4:1 , i need to check :
>>>
>>> You need to be really careful with this on a 6716. As you have hinted,
>>> you absolutely can blow away production traffic.
>>>
>>>>
>>>> *       On witch port it is recommended to connect the new probe
>>>> without affecting the port channel
>>>
>>> An unused port-group, or one with <2Gb/sec of TX traffic.
>>>
>>>> *       can one port on this module reach 8Gb of monitored traffic
>>>
>>> Yes, but only if the 4-port port-group has 8Gbit/sec of TX capacity.
>>>
>>>> *       the monitored traffic will be added to the actual Swicth load
>>>> or it is transparent ?
>>>
>>> Not sure what this means.
>>> _______________________________________________
>>> 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/
>>>
>>> _______________________________________________
>>> 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/
>>
>>
>> _______________________________________________
>> 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/
>>
>_______________________________________________
>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/