Hyper-V 2012 R Networking

[colly72]

Hi all,

 

We've just taken delivery of 2 Dell R630 servers, which i'm planning to use to virtualise our ageing physical servers. One server will be our primary host, replicating the vms to the secondary server. I'm trying to get my head round how we should set up the networking side of things. Each of the new servers has 2, 4x1GB NICs, so a total of 8 in each server. What's the best way to go about setting up the networking. Should I have 2 (minimum)for the management, plus another 2 or more for the VMs (both teamed for fault tolerance)? The servers have local storage, so there's no SAN to complicate matters. Do i just need to set up 1 virtual switch?

 

Any help/advice would be greatly appreciated.

 

Thanks.

[Eudora931]

Subscribe

[FN-GM]

Subscribe

 

You can subscribe using the thread tools without posting

[mikeyd101]

I'm no expert and very limited network teaming experience, but I'd team as many ports as I had free switch ports, every port you add will add available network bandwidth to the server. From what I understand about how network teaming works is that a single client can only use max of 1 ports bandwidth 1Gb.

 

What do you use management ports for? remote desktop? replication? Personally I wouldn't reserve ports strictly for management i'd just add them to the pool and monitor overall usage, if the teamed network interfaces were pegged then i might consider reserving some for management. If they were used for replication and the machines are close to each other i'd look at patching them directly into each other via cross over and remove the traffic from the switch.

[Steve21]

We have -

 

2x iSCSI

2x VM Traffic (e.g. external access)

2x Live Migration

1x CSV (Guess might be different if not a cluster)

1x Management

 

Steve

[supawat1]

Have you looked into Converged Networking for Hyper-V? You could team all 8 NICs and separating the team to have different vNICs purposes. https://www.youtube.com/watch?v=ed7HThAvp7o

[johnat1121]

I would suggest two 4gb LACP Teams to two separate switches for redundancy and then use vNICs to carve up the available bandwidth and -MinimumBandwidthWeight setting to ensure than in times of contention each service gets a set amount of bandwidth allocated and enabling SMB 3.0 in the hyper-v live migration settings will allow live migrations to make use of multiple NICs.

[3s-gtech]

My hosts have two onboard NICs with separate 4 port NICs on PCI-e. I have the onboard NICs doing host/management duties with the four ports teamed in Windows and tied to the vSwitch. Works well, with redundancy and speed.

[Eudora931]

FN-GM - would do, but doesn't work on phone app - tried several times :-(

[LytchettNM]

A couple of questions to try and understand what you are hoping to end up with.

Is the plan to set these new servers up as standard alone hosts and run VM's from each of them using the local storage or are you intending to cluster them to back each other up?

Why no SAN?

How many vServers are you hoping / need to run?

What's the spec of the new servers i.e. core's, ram and storage space?

What switching will they connect to?

Do you understand LAG / LACP?

Do you have core and edge switching in place and what speeds?

Do you have a routing device within your network you control?

Do you understand VLAN?

Lots of questions but I'm hoping this will allow for a better overview of what you can achieve with the kit you have.

And an idea of number of workstations and users could be useful.

 

Darren

[colly72]

Thanks for all the input. I've decided to go down the converged fabric route and team all 8 NICs together then create virtual NICs for live migration, CSV (if we decide to cluster later on), and management.