Hi All, I'm in the process of upgrading from standard to distributed switches on an ESXi 5.5 cluster. I notice on 5.5 that multiple LAGs (Link Aggregation Groups) are now supported - this has thrown a spanner in the works!
Traditionally as part of best practices I have always separated management traffic, vMotion and VM traffic into three separate standard switches with two physical uplink NICs each (I don't use FT or ISCSI). I was intending on replicating this when migrating to distributed switches and having three switches, however I notice that some people are now using one distributed switch with the uplinks divided into three LAGs which then send the relevant traffic to the relevant VMKernel or VM portgroups in that switch. Are there any advantages/disadvantages to either method and is one best practice over the other?
I think either one is effective as long as your upstream switches support multi-chassis ether channel like Cisco VSS or VPC.If you don't have multi-chassis etherchannel I would look at either running multiple switches as you planned on, or running a single switch and manipulating the active / standby / unused to separate traffic.
It's only from the management point. You can run all of them on one or on multiple switches. If you run all traffics on one switch you will make sure that the NICs that are used for one functionality is unused for another so you have the segregation. You don't have to worry about that in sperate switches but the management becomes complex with 3 switches to maintain. Why manage 3 when you can do with one. So the answer is "It depends". Both work just the same.
Thanks for the replies. I agree that there probably isn't one correct answer and it is dependent on infrastructure architecture. Because of the way our network is architected with multi-chassis ether channel and dedicated vmnics for each vSphere role, I think it will actually be easier (simpler for admins to interpret) to manage three separate role based switches. The VMKernel switches will remain static and unchanged, whilst the VM switch will be constantly evolving (new VLANs, portgroups etc).