Hi,
Thanks for this post that puts some light over CSV architecture!
First I write a few comments on the post and then please help me solve my below described challenge and understand a feature.
C1. The name "Cluster Shared Volumes" suggests this piece of technology is a shared volume i.e. shared block device but actually all the IO flows through the CSVFS. So naming this piece of technology a "cluster file system" is more suggestive than a cluster volume :)
C2. The post presents the 4 possible IO flows (Metadata, Direct IO, File System Redirected IO and Block Level Redirected IO) but does not say which mode is used by which MS technology like: Hyper-V, DPM or SQL 2014. This would be helpful for the storage planning and for setting realistic storage performance expectations for those technologies.
-----------------------------------------------------
The context of my challenge:
- We will run a Hyper-V cluster of 4 nodes (2012 R2)
- the SAN storage FC based and the storage controller is active/active supporting simultaneously IO to the same LUN via multiple paths and from multiple hosts
- we can present the storage resources (a pool of drives) to the Hyper-V cluster in one or several LUNs
Q1: From the CSV and Hyper-V cluster perspective (not the SAN perspective) is there any storage performance difference between presenting one large LUN to all the hosts or presenting multiple smaller LUNs to all the hosts and distributing the VMs even?
If I understood correctly this post and this one technet.microsoft.com/.../hh758090.aspx, Hyper-V 2012 and above makes always direct IO blogs.msdn.com/.../0268.DirectIO_5F00_Flow.png and all the hosts can make simultaneously IO to one or several CSVs and no SMB is interposed in this flow. Thus it does not matter if the pool if drives is presented to the Hyper-V hosts as one or several CSVs/LUNs.
-----------------------------------------------------
Some general questions regarding the CSVFS disk replication feature:
Vladimir Petter commented on 28 Apr 2014 11:46 AM that "CSVFS supports disk replication", please share more information about this topic.
Q2: Can CSVFS disk replication be used with the existing MS technologies or it is a feature (API) to be used by storage vendors?
Q3: Can CSVFS disk replication synchronous replicate running VMs (VHD/VHDX) between CSVs in the same or in different Hyper-V clusters? (I know Hyper-V replication but that one is asynchronous).
-----------------------------------------------------
Thank you,
Sergiu Pol.