I’ve been asking myself the question – what is the best way to provision Datastores using NetApp storage for my environment? And the answer is … it depends. It depends on a number of things both technical and non-technical. To answer my first question I must first figure out the answer to the following questions:
What kind of performance am I looking for?
In the realm of NetApp you can generally break this down into high-performance SAS disk and low-performance SATA disks … pretty cut and dry. Because of the way the WAFL file system works you get to take advantage of every spindle in the Aggregate for every LUN you create. Also RAID is taken care of because every Aggregate will be running RAID-DP by default which protects against double disk failures.
But wait … doesn’t the protocol impact performance?
Not so much these days –NFS, iSCSI, Fibre Channel are all pretty much the same in that regard. But don’t take my word for it, check out this presentation at VMworld 2012 by Vaughn Stewart and Chad Sakac!
What are the Technical limitations of VMFS-5?
- Min size 1.3GB
- Max size 64TB
- 128 ESXi hosts can connect at a time
- 256 VMFS datastores per host
- One VMFS datastore per LUN (recommended)
- Maximum number of files 130,000
Are there storage vendor specific features that need to be taken into account?
Yes! Volume size matters in NetApp. Some features of NetApp, such as Deduplication, have restrictions on what size volume they will work on. Volumes hold LUNs or are mapped as NFS volumes so if you want to use those features then you have to size accordingly.
So that 64TB datastore that you’ve always dreamed of may need to be reduced to 16TB depending on the version of ONTAP if you wanted to use deduplication and other features of NetApp. Each version of ONTAP has different restrictions so RTFM!
VMware Recommendations
For fewer, larger LUNS (datastores)
- You have larger datastores that allow you to create VMs without having the storage admin provision more space
- More flexibility for resizing virtual disks, doing snapshots, etc …
- Fewer VMFS datastores to manage
For more, smaller LUNs (Datastores)
- Waste less storage space
- Different applications need different RAID chararistics
- Remember with NetApp everything is RAID-DP
- More flexibility as the multipathing policy and disk shares are set per LUN
- Use of Microsoft Cluster Service requires that each cluster disk resource is in its own LUN
- Better performance because there is less contention for a single volume
Conclusion
When cutting up storage from NetApp Filers as LUNs I’m going with the Fewer Larger approach. For administration ease I’ll create Datastores that make management easier and let NetApp take care of the performance on the back end.