11 Responses

  1. Paul Meehan
    Paul Meehan at |

    Excellent visual and textual reference as always. I enjoyed the previous NFS posts too which are kinda related.

    You make the point above that “any” vmnic can be used for VM traffic in a LAG. That’s obviously true, but is it true for a VM communicating with a named other device with static IP address, that it will ALWAYS select the same vmnic based on the result of the hash ?. so it might always be vmnic0 for example ?. And if that fails then it will move to another active vmnic. And using IP hash is it possible for multiple uplinks participating in a LAG to become unbalanced, based on IP addressing schema, which might not happen with straightforward active/standby/route based on originating port id.

    Also is there ever a situation where more than a single port is simultaneously used (excepting iSCSI Multipathing) for any type of traffic ?. I am assuming it is not. So despite the presentation of 40Gb/s via 4 x 10Gb/s adapters in a LAG, to provide an example, no single object is only ever bound to a single vmnic at any point in time.

    So greater available bandwidth for a VM or for NFS would never one of the design criteria for LACP. Instead it would be mainly cost/lower latency or something to do with STP (maybe), as those traffic types will always be constrained by the bandwidth of a single uplink.


  2. Exploring Enhanced LACP Support with vSphere 5.5 | Wahl Network

    […] You can find my opinions on why you would and would not want to create a LAG, […]

  3. Drew Henning
    Drew Henning at |

    Great article Chris. You appear to be channeling your inner Frank Denneman with those drawings. 🙂

    Any chance you would share your thoughts on this for a Cisco UCS environment? I’m not exactly sure how it would look the FI’s already vPC’d to the upstream switches. Not to mention the vPC’s between the FI’s and the IOM’s and the IOM’s to the vNIC’s. (using the 2208XP’s)

    My initial thought is that the VDS + LACP config doesn’t make sense in a UCS environment since UCS is already doing this under the covers. But I’m not exactly sure…

    Also, I’d be curious to hear your thoughts on how you setup vNICs/QoS in UCS for VMware. I’ve done it the simple way with 2 vNIC’s/vNIC templates (A and B fabric), all vlans trunked, port groups sliced by vlan, and NIOC to try and keep everything happy. I’ve also done dedicated vNIC’s for VMware mgmt, vMotion, FT, IP storage, VM Traffic, etc. Then use UCS QoS policies to handle bandwidth and priority. Finally, I’ve been thinking about a hybrid config with only 2 vNIC’s, but the VDS passes QoS tags to UCS policies. I got back and forth with the pro’s and con’s of each approach. Tend to revert to the K.I.S.S. methodology.


  4. vSphere 5.5 Enhanced LACP Support Design Considerations

    […] In situations where you have more east-west traffic (between servers), LACP combined with MLAG is more efficient than any other load balancing mechanisms: for example, LACP with vPC on Cisco switches guarantees that VMs running on different hosts will always communicate through one switch. For more information and nice diagrams, please refer to Chris Wahl post, Revenge of the LAG: Networks in Paradise, […]

  5. Bjørn-T Nikolaisen (@btn003)

    Hi – Would not the identical path scenario in your second drawing (Red and gold) be a reason to always use LACP on VM-traffic, instead of using LBT and risk inter-switch traffic?

  6. Dell Force10 Part 2: VLT Basics | Niktips's Blog

    […] not required, but you may use this option later on if required. Read Chris Wahl’s blog post here for more […]

Share your point of view!