2009-03-23 05:49 AM
I'm planning the migration of our current HP EVA 4000 to a brand new FAS3140.
On the EVA i have 5 data volumes (on separated VDISK) for +/- 3TB available (+/-2.5TB in use before data cleanup). These volumes are presented to a MSCS acting as file server (1volume/ root share).
From migration point, the file server(s) are going to be replaced by the FAS3140. I'm wondering what is the best option in regards with building the new storage.
My thoughts so far:
Building a single 3 to4 TB volume and create the requested shares on this volume.
Vscan on on all the shares.
Is this conform to what would be the best pratices ?
I can't see much negative point for having one single volumes. We'll a dual-head config, volume will be splitted accross multiples shelves, and we'll keep enough spare disks to cover-up any kind of hardware failure.
any tips will be welcome !
2009-03-23 06:36 AM
On NetApp the aggregate is the base unit of storage which then contains volumes, which are logical items. The more physical disks the aggregate has the better the performance will be.
You have not said if you want to use LUNs or not.
You have said HA is something you are thinking about.
Have a read of this.
Hope this helps
2009-03-23 06:49 AM
Thansk for your answer.
I'll give some more details:
There will be +/- 380 users, using Office programs + Roaming profile and home dir
I'm planning to build one aggregate with +/-16 to 20 spindles (bases on SATA drives), on this aggregate I'm planning to build one volume that will be then access through CIFS. On this volume I'll then migrate the data currently held by the 5 LUN on the EVA.
I'm going to have LUN (FC in first place, iSCSI when the network will be upgraded and following HW replacement), but not for file services purpose.
SnapMirror is also on the list but that will come a bit after the file server migration.
2009-03-23 10:57 PM
Definitely towards fewer aggregates (single if possible generally speaking) as the more disks you have in an aggregate the better performance you get.
For multiple volumes vs. one, the factors out there involve....
2009-03-24 04:28 PM
Another option for moving data via CIFS...I just read about RichCopy in the April issue of TechNet magazine. There's also a link in the article for Robocopy GUI.
2009-03-31 02:55 PM
Don't forget VFM! VFM (Virtual File Manager) is a great tool for migrating data around, especially CIFS data (that's what it's best at really!).
Pretty much backing up what the other guys have said, keep the aggregates as big as possible to make full use of the spindle count, but create as many volumes as you need management groups or you need application granularity.Snapshots are taken at the volume level too, so you want to group data in a volume that you want the same retention and replication policies on.
For Exchange for instance you'd need several volumes to allow you to use SnapManager for Exchange. Whereas user home directories you could have one volume, and department shares could also be one volume but could be separated into Qtree's (think of them as management directories) for each department. You can replicate data (if you get that far in the design) based on either volumes or on Qtree's, so it pays to take a little time to plan the layout ahead of tiem.
Lots of options for you really, and the NetApp does a great job of helping you make the most of it.