I am new to Storage engineering so may not be asking questions the correct way. So bear with me. This is in context of deploying Vsphere 5.0 with Netapp FAS 6000 series storage. The esxi 5.0 hosts will run almost all kinds of vms ( windows , linux , database servers etc) . I have been reading some stuff NFS vs. iSCSI in a VMWare environment | The SAN Technologist, there is also one by netapp and vmware research .
The earlier post had some criticism on NFS based datastores, but the research paper shows that almost all protocols offer same kind of performance. So then how to decide on one over the other ?
Also can iSCSI or NFS based datastores be always given to the hypervisor first and then they are made available to guests or can guests have direct access to the same. Would appreciate some reference to read on this stuff.
For your data drives you will normally use raw device mappings (RDMs), where is no place for NFS - it has to be either iSCSI, or FC. There are some exceptions though: you can use the latest SnapManager for SQL in conjunction with databases & logs stored in VMDK files (on an NFS share) & a variety of Linux / UNIX clients / applications (e.g. Oracle) will happily use NFS mounts for storing data.
Typically you separate datastores presented to hypervisors from your RDMs and/or guest NFS mounts.