[c-nsp] mitigate output drops on Cisco 2960G platform

[Phil Mayers]

On 09/21/2013 12:08 PM, Phil Mayers wrote:
> On 09/21/2013 01:40 AM, Blake Dunlap wrote:
>> Making the linux box slower in the ways described (other than packet
>> pacing) would just increase the cpu load. If those are feeder servers,
>> you're much better off dropping the link speed to 100m or configuring TC
>> outbound than turning off a lot of the cpu offloading.
>
> If packet pacing is OK, then disabling TSO should be OK too surely,
> since they will result in similar results from a CPU load PoV - the CPU
> preparing the individual TCP segments, not the NIC.

Actually, re-reading this reminds me things are a bit more complex. As 
the current TCP pacing / FQ patches stand, they patch TSO to reduce or 
eliminate the bursting problem by sending sub-64k blocks in sizes 
adapting to the goodput rate. So assuming in future these patches are in 
use, TSO != pacing - the two complement.

(Whether the kernel or distros will change to use FQ or codel or 
something similar instead of pfifo_fast in future will be interesting to 
see, but the TSO sizing patch is not dependent on FQ)

Right *now* of course, TSO in Linux sends up to 64k super-frames to the 
NIC, and there's no such thing as pacing, so I stand by my assertion 
that disabling it will have an effect :o)

More info here, for the curious:

http://thread.gmane.org/gmane.linux.network/281264
http://thread.gmane.org/gmane.linux.network/281023