I currently have a single FAS2220 running 8.2.3 7mode, with 12x1Tb split in half and managed with dual controllers. Single volume of 1.5Tb on each controller, serving up VMs to ESXi hosts via NFS. I have de-duplication scheduled at quiet times weekly to avoid database maintenance, backups and peak usage - volume A Saturday at 2am, volume B Sunday at 6am.
I've noticed that the de-duplication time is getting quite large - 8 hours on volume A and 6 hours on volume B from last weekend, similar for at least the weekend or two before, based on my graphs of disk I/O. On the plus side, I am getting 47% savings on A and 55% on B from the de-duplication.
Are these times normal for a small 1.5Tb volume? Is it not incremental, i.e., it only has to scan changed blocks?
I know on a 12 disk array, I'm pretty tight on IOPS - I'm in the process of looking into either adding a shelf with another dozen disks, or replacing with a newer FAS25xx series 24 drive unit (just comparing costs) - mainly for the added spindles rather than capacity as the whole array is getting sluggish! How will that affect de-dupe time? Will the added capacity make it take longer to run, or the added spindles mean it runs faster?