[c-nsp] 7206VXR Big CPU Increase After 12.4 Upgrade

Mon Mar 26 09:46:16 EST 2007

That would explain it.  We're not doing any QoS on that router, yet.  I 
was going to start remarking packets this Spring when I do some other 
QoS work but it sounds like I should hold off on that a bit.

Thanks for the info, gents.
  Justin

Rodney Dunn wrote:
> If you are not using MQC for QOS or other features that would need
> to do packet classification like NAT or NBAR you probably will not
> see the increase that is being discussed.
> 
> See...we do make things better sometimes. ;)
> 
> On Mon, Mar 26, 2007 at 12:16:26AM -0500, Justin Shore wrote:
>> I thought I'd throw my experiences into the mix.  I have a 7206VXR w/ a 
>> NPE-G1 terminating about 800 PVCs of DSL customers from 4 OC3s.  I had 
>> to do an upgrade recently due to DHCP problems after an unexpected 
>> reboot.  I upgraded to 12.4(11)T1 Adv IP w/ LI.  The router is now 
>> consuming less CPU and less RAM.  The graphs seem to indicate that the 
>> internal temps may have dropped by a degree or 2.  My experiences don't 
>> seem to be following what others have reported.  I am curious to find 
>> out if my experience is simply a fluke since I have another identical 
>> router performing a different function that I need to upgrade in the 
>> very near future.
>>
>> Justin
>>
>>
>> Rodney Dunn wrote:
>>> Hank,
>>>
>>> I researched this for you and talked with the TAC SR owner.
>>>
>>> This performance hit is due to new packet classification mechanism
>>> that went in to 12.4(11)T.
>>>
>>> They are working to bring the performance back inline with the previous
>>> classification code. The only problem apparently is that it will be
>>> the next release of T as the changes are too large and complex to put
>>> in the throttle.
>>>
>>> I've asked them for some more clarification on what they are doing.
>>>
>>> I'll pass it on when I get it.
>>>
>>> Rodney
>>>
>>>
>>>
>>> On Wed, Mar 21, 2007 at 08:59:58AM +0200, Hank Nussbacher wrote:
>>>> Oldish thread:
>>>>
>>>>> Re: 7206VXR Big CPU Increase After 12.4 Upgrade
>>>>>
>>>>> For the benefit of the list and archives, we went back to 12.3.22 yesterday
>>>>> and our CPU levels dropped back to more like the previous levels.
>>>>>
>>>>> The "%IF-4-BACKWARD_COUNTERS" have also disappeared since we went back to
>>>>> 12.3.
>>>>>
>>>>> Would recommend staying with 12.3 if you don't need a specific feature of
>>>>> 12.4 as it is slightly more resource hungry (around 15 percent higher at
>>>>> just over 200Mbps in our experience).
>>>>>
>>>>> Thanks to everyone for their assistance,
>>>>> Mark.
>>>> Mark,
>>>>
>>>> We hit the same problem (CPU doubled - see graph from mid-Feb at - 
>>>> http://noc.ilan.net.il/stats/ILAN-CPU/tau-7206-cpu.html) going from 
>>>> 12.4(4)T1 to 12.4(11)T1 so if you want try 12.4(4)T1  and your CPU should 
>>>> not rise.   We have an open TAC case on this issue for 3 weeks now.
>>>>
>>>> -Hank Nussbacher
>>>> http://www.interall.co.il
>>>>
>>>>
>>>>> ----- Original Message -----
>>>>> From: "Rodney Dunn" <rodunn[at]cisco.com>
>>>>> To: "Mark Taylor" <maillist[at]smashie.ision.co.uk>
>>>>> Cc: <cisco-nsp[at]puck.nether.net>
>>>>> Sent: Tuesday, January 30, 2007 4:02 PM
>>>>> Subject: Re: [c-nsp] 7206VXR Big CPU Increase After 12.4 Upgrade
>>>>>
>>>>>
>>>>>> There are a lot of new features in the switching vectors between
>>>>>> 12.3 and 12.4 so you will see an increase in CPU.
>>>>>>
>>>>>> In an upcoming 12.4(last)T release that will then be 12.5 mainline
>>>>>> that is being changed so the net result will not be the same
>>>>>> CPU impact as features are enabled that has traditionally been the
>>>>>> case.
>>>>>>
>>>>>> Can you post your configuration minus any senstive information in it?
>>>>>>
>>>>>> Or unicast it to me...
>>>>>>
>>>>>> Rodney
>>>>>>
>>>>>> On Tue, Jan 30, 2007 at 02:01:39PM -0000, Mark Taylor wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> We upgraded one of our 7206VXR NPE-G1's from 12.3(12a) to 12.4(12) this
>>>>>>> morning to cover the recent security releases. We decided to make the
>>>>>>> major
>>>>>>> revision jump as we were reloading.
>>>>>>>
>>>>>>> After the reload the box jumped from 20 percent cpu to 35 percent on the
>>>>>>> new
>>>>>>> 12.4(12) image and at the time there was around 75Mbps in each direction
>>>>>>> of
>>>>>>> traffic running through the box - so an immediate increase of 15 percent.
>>>>>>>
>>>>>>> Now there is about 185Mbps each way and cpu is at 65 percent. This would
>>>>>>> have been about 45 percent before the change of image, so an increase of
>>>>>>> 20
>>>>>>> percent.
>>>>>>>
>>>>>>> This is looking very much like cpu could potentially increase more
>>>>>>> quickly
>>>>>>> than traffic. The jump in cpu between images is beginning to concern me a
>>>>>>> little. Has anybody else experienced this? Is it normal?
>>>>>>>
>>>>>>> The box is a peering router running IPv4 and OSPF, a full BGP table from
>>>>>>> one
>>>>>>> upstream and about 20 other direct peers at a small IX and a few ACLs.
>>>>>>> CEF
>>>>>>> is still running.
>>>>>>>
>>>>>>> I was hoping to get around 350Mbps through the box before running out of
>>>>>>> cpu
>>>>>>> and it was looking that way based on 12.3(12a). I guess I could go back
>>>>>>> to
>>>>>>> 12.3(22) which is fixed, assuming that the increase in cpu comes with
>>>>>>> 12.4.
>>>>>>> I just wanted to find out if this big jump in cpu between releases was
>>>>>>> normal/expected.
>>>>>>>
>>>>>>> Any help/experience appreciated.
>>>>>>> Thanks,
>>>>>>> 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/
>>> _______________________________________________
>>> 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/
>>>
>>>
> 
> 