Spiceheads,
I think I might be stuck...
The problem: Currently, users access a number of file servers each with a separate LUN off our SAN per share. Each file server is used to host say, 1 or 2 of these shares. (1 server, 1 OS vDisk, 2 LUNs). The team would like to have 2-4 file servers, that route traffic through a DFS-Namespace to these files without replicating the data locally. DFS-R accomplishes the first part of providing multiple points of access, but also requires multiple copies of data.
So I've looked for answers:
Clustered Share Volumes (CSVs) to my understanding would do the job, provided the servers are clustered, and running Server 2012 (as Server 2008 R2 had this feature, but only supported its use for Hyper-V). However, I also understand the data much be directly attached in that case, and accessible physically by at least 2 nodes in the cluster. The powers want the file servers to be VMs on ESX5.1 clusters.
New features of Server 2012 for "Continuous Availability File Windows Server" also come close from what I can see of the "classic" configuration, but I don't see Microsoft explicitly stating that the "One LUN, one node accessing at a time" part has been overcome. Also, I understand this solution adds a bit of overhead (can be significant due to the size of our shares) and may not support Data De-Dup, File Quotas, etc.
I know this has been commonly overcome before using third party vendors with special hardware on the SAN side to overcome the NTFS "shared nothing" problem with custom file systems. Unfortunately they are not available here.
There is the option of creating a 1:1 file-server-to-site setup here. We have two locations and the data is already being replicated to the other location. Bandwidth to the second site is of course the limiting factor there.
So my question is, has anyone else successfully accomplished this task of creating a highly available file service, without multiple copies of the data at the same site?
Thanks in advance!