VMware Cloud Community
markw78
Contributor
Contributor

ZFS on NFS or iSCSI VMFS

iSCSI allows me to tie LUN's to IP's which has advantages, though may be able to tie NFS shares to IP's and get the same effect...

NFS would allow me to mount the storage on the physical server it's attached to, as near as I can tell I can't use net-snmp to monitor the unmounted iSCSI ZFS LUN's or Zpools. I'd like to monitor the LUN's at the SAN level as well as ESX Lun, Data Store and Guest OS perspectives.

NFS lets us work with the VM files a lot easier, we can copy them locally or off to another environment for testing or whatnot.. This is nice, then again we have Windows PC's in a Windows shop that won't nativly mount NFS, which brings me to the next point, a 1up for iSCSI...

We don't have any Linux admins in the company so much, some who know a lot but thats it. There's a benifit to leaving the overhead and unknown of NFS out of the picture.

I'm leaning twords iSCSI for supportability, I'd imagine we can configure LACP on the Solaris box to get Xgbps and do the same with load balancing on the ESX hosts... Spindle count and config is a whole other ball game!!

Anyone with opinions to any of this? Links to docs etc? have been searching and reading, there's too many options!

Reply
0 Kudos
1 Reply
markw78
Contributor
Contributor

And also, I noticed there is an option for Disk Share cost, would this allow me to keep the stoage more general IOPS avaialble, single raid set... and let ESX divy up access to it? Not sure how disk bandwidth access policy type thing would effect small random io like DB or the like.

Obviously at some point I will exceed the capable IO of the spindles, can ESX help me prioritize those IO's and cripple the Guest VM's of my choice, rather then whichever happens to get access to the disk?

Reply
0 Kudos