Server Message Block Usage
While Server Message Block (SMB) has been available in Windows for a long time, its
usage has been limited to basic file-sharing scenarios, such as users accessing their
home drives or a file share containing archived data. Hyper-V had long had the
requirement of having block-level access to its storage; that is, the host mounted the
volumes that contained the virtual machines, which could be direct-attached or
connected via mediums such as iSCSI or Fibre Channel. However, this was a challenge
for many organizations that were used to using file-level protocols with virtualization.
Specifically, VMware supported NFS for virtual machine storage, which was available
in many NAS solutions that typically are much cheaper than SAN solutions and are a
good fit for many environments.
Windows Server 2012 invested greatly in SMB to make it an enterprise-ready solution
suitable for storing virtual machines and other enterprise workloads, such as SQL
Server databases. SMB 3 introduced many new features and performance
improvements to make it a realistic choice for virtualization storage.
SMB Technologies
I previously talked about SMB being used to store user documents, and now with SMB
3, it will be used to store mission-critical virtual machines. This requires a big shift in
resiliency and failover technologies.
When the user is editing a PowerPoint document from an SMB share, portions of the
document are cached locally, and occasionally the user clicks Save. If the SMB file
server experiences a problem—for example, if it reboots or if it’s clustered and the file
share is moved to another node in the cluster—the user would lose the handle and
lock to the file, but that really does not have any impact. The next time the user clicks
Save, everything is reestablished, and no harm is done.
Now consider Hyper-V storing a virtual machine on an SMB file share that
experiences a problem, and the file share moves to another node in the cluster. First,
the Hyper-V box will wait for the TCP time-out before realizing the original
connection has gone, which can mean 30 seconds of pause to the VM, but also Hyper-
V has now lost its handles and locks on the VHD, which is a major problem. Whereas
user documents may be used for a few hours, enterprise services like a virtual
machine or database expect handles on files to be available for months without
interruption.
SMB TRANSPARENT FAILOVER
Typically, for a clustered file service, a single node of the cluster mounts the Logical
Unit Number (LUN) containing the filesystem being shared and offers the share to
SMB clients. If that node fails, another node in the cluster mounts the LUN and offers
the file share, but the SMB client would lose their handles and locks. SMB