I know, I know... really I should just get on with server core and do the job properly. In my more conspiratorial moments I have suspected that this is Microsoft's intention, and that server core itself is a plan to boost the numbers taking MS certification.
Just to give you some quick results from some fileservers - just because some beat the 28%. Need to do some CPD on the need not to copy from the shared area to my documents! Obviously big advantages up front - has anyone come across any downsides?
Processed files: 331578
Processed files size: 539.95 GB
Optimized files size: 328.72 GB
Space savings: 211.17 GB
Space savings percent: 39
Processed files: 117387
Processed files size: 200.52 GB
Optimized files size: 126.07 GB
Space savings: 74.43 GB
Space savings percent: 37
Processed files: 150567
Processed files size: 298.56 GB
Optimized files size: 223.50 GB
Space savings: 75.05 GB
Space savings percent: 25
I'm not sure if it is a downside or not, we are looking at putting all our file shares on one drive on one server to get the most out of deduplication. We run a hyper-v environment so really this shouldn't have any performance impact compared to on 2 different virtual servers and 4 different virtual drives at the moment.
Just feels like adding a single point of failure, but really it isn't I guess!
Just poked around on one of ours (2008R2, running SIS) and it's quite effective.
There is a downside to deduplication, and to SIS. If a sector becomes corrupt for whatever reason, and it happens to be a file (SIS) or 'chunk' of a file (Deduplication) that has multiple copies across the disk, you end up losing all those copies.
But then, we all have backups!
25% is about right for user files according to Microsoft's guides. Up to 50% for shared areas. 75% for software deployment areas and 90% for VHDs.
I have enabled this in production now. But when i look at the disks in Server Manager i can see a deduplication rate of 30% i right click and refresh it changes to say 28%. I refresh it again and it goes back up to 31%. Why is it so erratic, changing the figures in seconds? Will it settle down?
The process is done via a pair of scheduled tasks each night. I believe they have a limit of how long they run for each night also. So, yes, the values can go up over time as it optimises more and more.
I have come across a slight issue with backups. Backing up using BackupAssist seems to have issues with the hard links that are used, and the backups often just end up as a bunch of 4k files rather than the actual files.
Our Veeam setup makes great use of dedupe for the backups. I forget the ratio now but it helps a lot, not just for storage but also (because this is source dedupe) in terms of how much data we have to transfer over the wire for our offsite backups to another school a few miles up the road.