vSAN ESA NIC possibilities

[#1]

Hi,

when looking at "vSAN ESA ReadyNode Hardware Guidance" (https://www.vmware.com/resources/compatibility/vsanesa_profile.php), there are 2 of the configurations with 2x 25Gb NICs, vSAN-ESA-AF-6 and vSAN-ESA-AF-HighDensity.

In: "Designing vSAN Networks - 2022 Edition vSAN ESA" (https://core.vmware.com/blog/designing-vsan-networks-2022-edition-vsan-esa), it is stated "vSAN ESA AF6, and AF High Density nodes require 50Gbps of vSAN networking throughput", which seems to fit with the 2x 25Gb.

In: "Designing the vSAN Network" (https://docs.vmware.com/en/VMware-vSphere/8.0/vsan-planning/GUID-031F9637-EE29-4684-8644-7A93B9FD8D7...), it is stated: "vSAN does not use NIC teaming for load balancing." and "vSAN supports IP-hash load balancing, but cannot guarantee improvement in performance for all configurations. You can benefit from IP hash when vSAN is among its many consumers. In this case, IP hash performs load balancing. If vSAN is the only consumer, you might observe no improvement. This behavior specifically applies to 1-GbE environments. For example, if you use four 1-GbE physical adapters with IP hash for vSAN, you might not be able to use more than 1 Gbps. This behavior also applies to all NIC teaming policies that VMware supports.". 

The last document "Designing the vSAN Network" seems to somehow contradict my understanding of the other documents statements of using 2x 25 Gb NICs to gain 50Gbps of vSAN networking throughput.

So question is, if you can get benefit of 2x 25Gb NIC teaming for vSAN vmkernel communication alone? The other understanding would be, that you can benefit of 2x 25Gb NICs by dedicate one NIC alone for vSAN vmkernel communication (and all other communication through the other NIC), but then the statement "50Gbps of vSAN networking throughput." would be misconfusing. 

Best regards

Henrik

[Lalegre]

@HenrikPersson,

Yes it is explained with different words on every document so it is a little bit confusing. Basically the only way you can see quite better performance with load balancing is by using IP Hash configurations but this will not be 100% of the cases. This is because vSAN uses a single VMkernel and it does not sees multiple IP Hashes to root as it would be a portgroups with many virtual machines.

Now regarding the requirement, the minimum needed is 25Gbps as you clearly mention and you could dedicate one NIC for vSAN traffic and have a second one as Standby in case of a failure and simply use Route based on originating virtual port. 

Always is recommended to have two adapters in case of whatever failure it could happen.

[depping]

I have asked the vSAN team to re-word some of the documents. I had a discussion with them and LACP (etc) doesn't indeed lead to 2x25=50Gbps, the are probably seeing and additional 40/50% of bandwidth at this stage. Which is a lot better than in the past I have to say, but still not what some documents may suggest.