Scale-Out File Server Question

Hello

I am in the process of building a Windows Server 2012 Scale-Out File Server. I have been following the steps in Jose Barreto’s Blog here: http://blogs.technet.com/b/josebda/archive/2012/08/23/windows-server-2012-scale-out-file-server-for-sql-server-2012-step-by-step-installation.aspx

There are a few differences in my environment.  Firstly, I am building this with physical servers (not VMs). Secondly, my goal is to use this Scale-Out File Server for Hyper-V (not SQL).

This is a fairly long and complex environment, but I’ll try to sum it up as quickly as possible. I have two physical servers connected to a shared storage (FS1 & FS2). These two servers are now running Windows Server 2012 RTM.  They are in a cluster with 4 separate LUNs configured as Cluster Shared Volumes (CSV1, CSV2, CSV3 & CSV4). The cluster is running the Scale-Out File Server Role (SOFS).  So far so good, it all appears to work quite well.

Next, I’ve installed Windows Server 2012 on my four “soon to be” Hyper-V servers (HV1, HV2, HV3 & HV4). I say “soon to be” because I haven’t actually installed the Hyper-V role on any of them yet. Before doing so, I figured I would simply play around with the Scale-Out File Server and see how it works.

These servers are connected to a few different network VLANs.

1. SERVER: The SERVER network is where all of our Windows Servers connect in order to talk to each other (ex AD, Exchange, etc…)  And this is obviously the network that the clients workstations get routed to in order to access resources on these servers.
2. HA: The HA network is a network that is to be used strictly for the Failover Clustering traffic.
3. STORAGE: The STORAGE network is used by the Scale-Out File Server and Hyper-V servers only. This should be the preferred network to use by the Hyper-V servers for accessing the VM files stored on the Scale-Out File Server. Each of the two File Server nodes have four 1Gb NICs in this network which are TEAMed together to form a 4Gb connection. Each of the four Hyper-V Servers have two 1Gb NICs in this network which are TEAMed together to form a 2Gb connection.

I’ve been simply doing some testing.  My tests are simply to copy a large file (80GB) from the SOFS to my Hyper’V Server’s C: drive, and monitoring what is going on. Here’s what I’m seeing:

1. From HV1, I copy a file from the SOFS’s CSV1.  On HV1, the STORAGE nic is receiving @ 1Gbps. On FS1, no significant network activity.  On FS2, the STORAGE nic is sending @ 1Gbps. Cool. So far so good.
2. From HV2, I copy a file from the SOFS’s CSV2.  On HV2, the STORAGE nic is receiving @ 1Gbps. On FS1, the STORAGE nic is sending @ 1Gbps.  On FS2, no significant network activity.
3. I repeated the test on HV3 and HV4 with the exact same result as HV2.

Now, here’s my question.  Why is FS1 sending the file to HV2, HV3 and HV4, and FS2 sending to HV1? The SOFS role is running on FS2.  The CSVs are evenly divided between FS1 and FS2. So how does it decide which node it talks to?  My concern is the following. I want the Scale-Out File Server cluster to be Active-Active and to split up the workload as evenly as possible. How do I make sure to accomplish this?

Sorry for the extremely long post.