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

[Phil Mayers]

On 20/09/2013 17:56, Martin T wrote:
> Hi,
>
> I have three WS-C2960G-24TC-L switches which all have about ten
> servers connected. Each switch has a single 1GigE uplink. Those 1GigE
> uplink ports on all switches experience occasional OutDiscards.

As per Gert's reply - this is a platform "feature". Even small 
microbursts can overload these devices.

Unless you can shape the link-layer egress behaviour of your hosts, you 
will have to live with this, provision more uplink bandwidth or change 
device. See the archives for upgrade suggestions, but be aware that Cat 
3560/3750 have the same issue; you need to go to 4500/6500 or 
Nexus/DC-grade switches before it improves.

(However: I wonder if the various techniques being discussed elsewhere - 
like the new Linux kernel packet pacing, disabling TCP segmentation 
offload and reducing nic ring, kernel and app buffers - might alleviate 
the generation of microbursts?)