Testing Windows Server and the Scale-Out File Server – What should your lab look like?
With the release of Windows Server Technical Preview, a common question comes to the surface again. What kind of lab hardware do I need to play with it?
Ideally, you would get one of those new Cloud Platform System (CPS) racks with a 32-node Hyper-V cluster, a 4-node Scale-Out File Server and 4 JBODs with 60 disks each. However, not everyone can afford a lab like that :-).
I am recycling some of my previous step-by-step guides and offer some basic options on how to test the Windows Server 2012 R2 and Windows Server Technical Preview features.
IMPORTANT NOTE: Some of these configurations are not supported and are not recommended for production deployments. Please see each linked blog post for details.
Option 1 – Virtual Machines
You can do a lot with just a single physical machine running Hyper-V. With enough RAM and a nice Core i7 CPU you can configure half a dozen VMs and try out many scenarios.
This would be enough to to test Failover Clustering, Scale-Out File Server, Shared VHDX, Scale-Out Rebalancing and Storage Replica. You can even test some basic Storage Spaces and Storage QoS capabilities.
Obviously, you can’t do much testing of Hyper-V itself and certainly no Hyper-V clustering features. This will also not showcase storage performance at all, since you’re limited by hardware.
This is the setup I used many times on my beefy laptop with 32GB of RAM and used for many demos in the past.
Option 2 – Physical Machines, basic networking
If you’re testing new Hyper-V features related to clustering, you can’t run virtualized. This would require at least two physical Hyper-V hosts.
Also, might want to test how Storage Spaces works with a physical JBOD plus a mix of HDD/SSD to test storage performance, tiering, etc.
A setup like this will be ideal to test Hyper-V high availability, Storage Spaces performance and the full capabilities of the new Storage QoS.
To do some basic testing you don't need high-end hardware. I've done a lot of testing with basic (even desktop-class) hardware. You do want to make they properly support Hyper-V.
While the setup in the link above suggests the use of RDMA, this will work fine with regular networking, even 1GbE. It will obviously perform accordingly.
Option 3 – Physical Machines, RDMA networking, all SSD
You can go one step further if you add RDMA networking to you setup. This means you need some RDMA cards and a more expensive switch. I might be a little noisier as well.
You might even go as far as to have multiple RDMA NICs so you can try out SMB Multichannel at high speeds. At this point you might want to move to an all-SSD configuration.
At that point you want to make sure the hardware for both the Hyper-V host and the File Server are server-class equipment with good performance, fast CPUs, lots of RAM.
This is probably the simplest configuration to showcase extremely high performance. Note that the example above does not include failover clustering.
Option 4 – CPS
Beyond that, you are now in CPS territory, which includes multiple Hyper-V hosts, multiple file servers, dual RDMA paths, multiple JBODs, tiering, the whole thing.
Details at http://www.microsoft.com/cps
Bonus – Option 5 – Azure VMs
An interesting option is to skip the physical lab completely and go all Azure with your testing lab.
It's fairly easy to configure an Azure VM as a standalone file server and use Azure data disks to experiment with Storage Spaces. You can also try out Storage Replica.
With a little care on how you configure your Azure networking, you can also setup a Scale-Out File Server using iSCSI for your shared storage.
I hope that gives you a few ideas on how to get started with your testing.