I am trying to find the information I need to set a couple of servers up for Hyper-v and SCVMM 2012 SP1.
We are a small company having approximately 10 hyper-v. I have installed SCVMM 2012 SP1 and want to make it the right way the first time so I have some questions about NICS and how to set them up the right way.
I want to set up a couple of dell Power edge R720 (Win 2012) that we are using in the dev department as Hyper-v servers with SCVMM on top. The library server resists on a Windows 2012 NAS.
The storage is on the local machines so no cluster. Both machines have 4 NICS that we wants to utilize the best way possible.
A couple of clients on one of the servers are internal production servers, like Team Foundation Server and sql server, but non of them are utilizing there hosts very much.
These servers are hosting test clients as well. If a developer or tester needs a place to test our software he can request a template from the library server and let it spin up. When he is finished he will delete the server until next time. Normally he will RDS into the server and work with it.
How should I utilize the NICS in this case.
I have one NIC that are reserved for remote desktop into the host, with a static ip and computer name registered in DNS.
The rest of the network cards are teamed with switch independent, Hyper-v Port load balancing.
I have checked that our switch is supporting LACP, so it could be nice if we could use some of the bandwidth for the connection between the library server and the host, when copying test and demo systems from the library. But are the connection between the library server and the host using the first NIC or the Teamed Ones? My own best guess would be the first one.
Would it then be a good idea to make two teams with two NICS in each and then use LACP to make the transfer between the library server and the host as fast as possible?
This was a long text for at “simple” question, but I hope you understand?
Host networking is one of those things that can be done in numerous different ways and they are all applicable, supported etc but there's little 'advice' out there on which way is best.
You are correct that your host will communicate with VMM &/or the Library server using the single Host NIC you have configured. Creating multiple teams is supported and would provide redundancy for your host connection also.
Alternatively, you could use the converged network/fabric approach. This is a little more complex to configure at first but it can be much easier to manage in the long term. Converged fabric involves creating a single Team using all 4 of your NICs, you then create vNics on the host for each purpose like Management, Live migration, CSV etc
This would give you the ability to manage QoS and prioritise traffic also.
What speed of Nic are you usingin the hosts?
Do you find that your existing 1 Nic for management is being saturated?
What speed of Nic is in your NAS?
You may find that your NIC's are not holding you back at all, in which case there's not much to worry about.
This is a good article on how to do the converged fabric via powershell or via VMM.