topic VMWare and Netapp using FCP recommendations in VMware Solutions Discussions

VMWare and Netapp using FCP recommendations

CHRISLAWLOR — Thu, 05 Jun 2025 06:32:40 GMT

hi all

we have a customer who has bought VMWare (vsphere 5) and a FAS 3240C - they have the FCP license and wish to present LUNS to the 3 ESXi hosts with around 3TB for virtual machines

what would you recommend to do in terms of datastores and LUN presentation ? we were thinking to present 5 x 600GB LUNs to all 3 ESXi hosts, and create a seperate datastore per LUN, each LUN would also be in a seperate volume on the same aggregate

im aware that we would lose the dedupe capability between the datastores with the above

any suggestions would be great

thanks

Re: VMWare and Netapp using FCP recommendations

radek_kubka — Wed, 14 Mar 2012 10:25:38 GMT

Hi and welcome to the Community!

Datastore size is a subject to a personal preference, but I would probably lean towards fewer & larger datastores (still sticking to the rule one datastore, one LUN per volume).

LUN-wide locking (leading to the idea of having more smaller LUNs) was a bit over-hyped in vSphere 4 & in version 5 this can be safely ignored completely thanks to the introduction of Atomic Test & Set (ATS), http://blogs.vmware.com/vsphere/2011/07/new-vsphere-50-storage-features-part-1-vmfs-5.html

From the NetApp side perspective this is also worth noting - http://media.netapp.com/documents/tr-3749.pdf, p9:

Although spanned datastores can overcome the 2TB LUN size limit, they are most commonly used to overcome scaling limits imposed by storage arrays that use per LUN I/O queues. These traditional architectures place a limit on the number of simultaneous commands that can be sent to a LUN at a given time, which can impede overall I/O performance in a datastore.

Storage arrays powered by Data ONTAP are free of this legacy architecture; therefore, NetApp does not recommend the use of spanned VMFS datastores.

Regards,

Radek

VMWare and Netapp using FCP recommendations

CHRISLAWLOR — Tue, 27 Mar 2012 09:12:20 GMT

Thanks Radek

Chris