Waaaaay too many people have tried to test SOFS performance (and therefore SMB 3.0 as the medium and often Storage Spaces as the backend storage) by copying files and watching transfer rates. I have news for you: this is not what SMB 3.0 was intended for. File copies are chatty and inefficient. SMB 3.0 is for application data such as data streaming to/from a Hyper-V virtual machine or SQL database files.
I know that Microsoft got hit with this over and over and over and over. I know I’ve seen it on various social media types over and over and over and over and over. And people don’t listen or read. So Microsoft has a KB article.
Consider the following scenario:
- You use Storage Spaces to create storage pools on a Windows Server 2012 or Windows Server 2012 R2-based failover cluster.
- You have a file server or scale-out file server resource that hosts continuously available Server Message Block (SMB) 3.0 (or a later version) shares that are backed by the storage pools.
- You try to copy large files from a Windows 8 or Windows Server 2012 computer to the continuously available SMB shares.
In this scenario, the observed copy throughput is significantly lower than what the connected network and storage system can support.
The fix is included in KB2934016, the big “Update” for Windows Server 2012 (and Windows 8 and Windows RT).