During a recent Q&A session regarding a HP Blade implementation (BL460G6 & 490G6) much confusion was had surrounding the presentation of NICs to the individual Blade servers, how the host operating system would see them and what the throughput would be.

Here’s a very brief summary of how things are:
  • Each server has 2 x 10Gb LOMs (LAN on motherboard). Conceptually I find it easier to think of this as having a single switch on either side of the rack <- Something you would normally do in standard rack server implentations.
  • Each LOM is divided into 4 x NICs via a PCIe function of Flex-10. Think of it as having 2 x Quad Port NICs installed which means the operating system will see 8 x NICs. These NICs are referred to as Flex-NICs. It’s worth mentioning here that 8 x NICs are the maximum, you can have less.
  • Balance across the Flex-NICs are you would normally do taking into considering resilience by teaming the networks across both LOM1 and LOM2.
  • Flex-10 configuration permits different bandwidth presentation to each Flex-NIC, the increments for allocation are in 100m/bit.
Caveats
  • You cannot present the same VLANs across 2 different Flex-NICs on the same LOM.
  • The sum of bandwidth must not exceed 10Gb per LOM.
  • Maximum VLANs per Flex-NIC cannot (currently) exceed 28.

This may be of help when planning your next VMware implementation.

[Many thanks to Alex H for his insight and simplistic explanations]