ONTAP Discussions

LUN - Dedupe - Thin Provision or not ?

NETAPPORATI
4,932 Views

Hi All,

We have a FAS 3210 and 6TB of Usable Storage, I am now evaluating / planning on what type of LUN shall I create, I have created an aggregate spanning all disks assigned to ONE controller and then created a volume on top of it, so far so good. Now this LUN is going to be used by a Windows 2008R2 host for storing a folder structure with more then 30,000 folders and a lot many folders and text files within these 30,000+ folders. At present this data is around 1.2 TB and growing at 2GB - 3GB per 24 hours. I have been reading Dr.Dedupe articles (very informative and interesting) and this is what I have concluded so far, just seeking your expert suggestions / re-assurance :

a) I will create a LUN (Thin Provisioned) with no Space Guarantee, then mount it on Windows 2008R2 host, copy all the data over, run a manual FULL Dedupe on the VOLUME hosting this LUN ? RIGHT ? this I think will allow me to save space at Volume Level. I have set "0" fractional reserve at Volume Level, inside this Volume I have two LUNS created at the moment with space guarantee.

b) When I select volume (netapp system manager - mmc based) and click Edit, on the last tab of the property sheet, I see I can specify maximum number of files and a maximum directory size, I think this is related to WAFL and doesn't relate to a Directory created inside the LUN ?

I have read the following blog article from Dr.Dedupe and I hope I have understood it correctly 🙂

https://communities.netapp.com/community/netapp-blogs/drdedupe/blog/2010/05/26/luns-deduplication-and-thin-provisioning--the-three-amigos-of-netapp-un...

Enter Thin Provisioning.  If the LUN space guarantee is turned off, the storage system still tells the server that the LUN is there but never actually carves out the space (see I told you LUNs and servers were dumb).  At this point the LUN is thin provisioned.  No space is consumed on the system until data actually starts coming into the LUN, i.e. space is only provisioned as needed.  Better yet, when the LUN is dedupe'd - all the reclaimed space goes back to the Aggregate (remember those?) where it can be re-used by any other FelxVol or LUN within the Aggregate.  Cool, huh?

Will be grateful for your expert suggestions and insight.

Thanks !

1 ACCEPTED SOLUTION

scottgelb
4,932 Views

You shouldn't have to recreate the volume.. you can change volume guarantee on the fly with vol options.  You just have to decide how you are going to monitor space so that you don't run out of disk on the host...setting up snap autodelete and/or volume autogrow along with monitoring.

Having 2 luns in the same volume will help dedup rates if their data is simlar since dedup is per volume.  If the snapshot schedule is the same (if you run snaps) then it makes sense to keep them in the same volume container.

With lun clone, clone or flexclone you can clone the entire volume or lun without any space overhead regardless of one or two volumes.

Typically we don't see space guarantee anymore (some still keep some) and just size for what is needed with some extra safety net.  Assuming a LUN will fill 100% and if you get 20% dedup rates means you need at least 80% of the lun size in the volume/aggr available plus snapshot deltas for example...depending on where guarantees are set.

View solution in original post

4 REPLIES 4

scottgelb
4,932 Views

Very good summary.  You are correct that maxdirsize and maxfiles are more for NAS and don't apply to the LUN since it is objects in a directory/volume and a LUN is an individual object.

For dedup you are also right... thin provision the volume and the space saved on dedup goes back to the volume..  If you also thin provision the volume that space saved goes back to the containing aggregate.

NETAPPORATI
4,932 Views

Hi Scot Gelb,

Thank you for your reply and assurance, I have already created the volume ( I can still afford to delete and re-create it), do you think I should delete and re-create a THIN provisioned volume?

a) IF it is good to create a THIN provisioned volume, are there any recommendations / best practices I should consider / follow while creating the VOLUME inside the aggregate ?

b) As There will only be TWO LUNS at max inside this volume for these two servers only, can I just get away with creating THIN provisioned LUNS now? I have copied data so I would have to run the copy process (11 Hours) again ? The aggregate will only have ONE Volume on this or do you see any advantage of creating two thin provision volumes ?

c) By creating two volumes, can i restore snapshots of one volume into another, I mean can I mount them (snapshots) inside another volume or by creating THIN provisioned volumes I would have the opportunity to use the unused space to restore snapshots ? As of now, when I click on a snapshot (volume snapshot), it doesn't give me the option to restore or to view files inside it?

d) Do I create a thin provision volume with space guarantee and auto-grow enabled ?

Sorry to have bombarded with 4 more questions, just want to be sure, I am still learning and learning !!m 🙂


Thanks so much for your response !!

scottgelb
4,933 Views

You shouldn't have to recreate the volume.. you can change volume guarantee on the fly with vol options.  You just have to decide how you are going to monitor space so that you don't run out of disk on the host...setting up snap autodelete and/or volume autogrow along with monitoring.

Having 2 luns in the same volume will help dedup rates if their data is simlar since dedup is per volume.  If the snapshot schedule is the same (if you run snaps) then it makes sense to keep them in the same volume container.

With lun clone, clone or flexclone you can clone the entire volume or lun without any space overhead regardless of one or two volumes.

Typically we don't see space guarantee anymore (some still keep some) and just size for what is needed with some extra safety net.  Assuming a LUN will fill 100% and if you get 20% dedup rates means you need at least 80% of the lun size in the volume/aggr available plus snapshot deltas for example...depending on where guarantees are set.

aborzenkov
4,932 Views

Another typical case when I would recommend placing multiple LUNs on the same volume is to get "crash consistent" snapshot of these LUNs. E.g. for all LUNs belonging to the same server or hosting parts of the same database.

Public