[c-nsp] CSR1000V and CPU usage

Robert Hass robhass at gmail.com
Thu Aug 13 07:34:36 EDT 2015

> We use the CSR1000V on ESX as well.

Can you show me how much CPU usage you currently has ?

show processes cpu platform sorted | exclude 0%      0%      0%
show platform software status control-processor
show platform hardware qfp active datapath utilization
sh int | inc rate

What features you're using on this CSR and what resources are allocated to


On Wed, Aug 12, 2015 at 11:37 PM, Pshem Kowalczyk <pshem.k at gmail.com> wrote:

> Hi,
> We use the CSR1000V on ESX as well. From my experience - the code that
> calculates the load of the router is most likely not aware it runs within a
> VM so the calculation are done in relation to 100% CPU utilisation, but
> since the number of allocated cycles might change the resulting number is
> relative.  What ESX reports is probably closer to truth (but also take it
> with a grain of salt). In our tests we pushed over 2.5Gb/s through a single
> instance of CSR1000V and over 500k pps (AX licence, with pinned resources)
> for prolonged periods of time with no problems.
> I do not think you should worry about that process.
> kind regards
> Pshem
> On Thu, 13 Aug 2015 at 08:35 Roland Dobbins <rdobbins at arbor.net> wrote:
>> On 13 Aug 2015, at 1:24, Robert Hass wrote:
>> > I deployed Cisco CSR 1000V as edge router in DataCenter.
>> Deploying any variety of software-based router at one's edges is a
>> mistake, and has been for many years.
>> The Cisco virtual stuff is great for labs, training, testing, and so
>> forth - kudos to them for producing it, and I hope they do even more
>> with their virtual versions.
>> That being said, there's no way I'd deploy any of it to route actual
>> packets on actual production networks.  Nothing against Cisco nor their
>> virtual stuff, but in any kind of Internet-facing environment,
>> software-only doesn't scale.
>> At some point in the future, this will change, as hardware-based
>> routers/switches/whatnots will take the 'nFV' trend even further, and
>> software hypervisor-based ones will gain direct, high-performance access
>> to serious hardware-based NICs, NPUs, et. al.  But for now, I personally
>> think it's way too soon to be doing this in production environments.
>> -----------------------------------
>> Roland Dobbins <rdobbins at arbor.net>
>> _______________________________________________
>> 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/

More information about the cisco-nsp mailing list