6 Responses

  1. joakim
    joakim at |

    Take a look at the discussion on ietf-http-wg about the upcoming HTTP/2 spec and see that jumbo sized MTUs are likely to be important in an even more global sense in the future.

    Reply
  2. @vcdxnz001
    @vcdxnz001 at |

    One of the problems being faced now at least in the vSphere arena is that NSX for vSphere being tightly coupled to the vSphere kernel means that it can’t handle fragmentation at all. The vSphere Kernel doesn’t do fragmentation gracefully, and there have been quite a few bugs with the network stack recently. My guess is the QA on Jumbo in the stack is quite limited. So this makes it critical that the MTU configuration of the underlay network is set correctly so as to not cause havoc with the vSphere Kernel when using NSX or vCNS with VXLAN and other overlay protocols.

    Reply
  3. Newsletter: July 27, 2014 | Notes from MWhite

    […] Jumbo Sized MTUs are coming! Some good background on Jumbo frames both now and in the future.  Good […]

  4. Sunil S
    Sunil S at |

    not worth the aggravation and configurations for a few percentage points of extra performance. 99% of time it is unnecessary. in the datacenter KISS reigns. hardware is set at defaults for a reason, always try to avoid messing with defaults.

    Reply

Share your point of view!