Re: Note to IBM
I used to work at IBM, and am now at a competitor. The 900 is a good product as was the 820 before it (forgetting about the non-existent RAS for a moment). It's the nichest of niche, as all flash storage is (despite what they tell you, you're throwing money away if you're putting data on flash that doesn't need to be there). It was designed to be as low latency as is possible, and that meant a lack of features and a lack of configuration options.
IBM bolted on XIV and SVC, as two attempts to provide these features. This does the job, but takes the 900 away from what it was intended for and into the realm of the competition, most of which does a better job.
These changes are nothing exciting apart from the potential to reduce cost, although it'll still be expensive unless you absolutely need very low latency on all the data you're putting on it.
As for dedup, and compression for that matter, these have one purpose, and one only: potentially reducing cost. They serve no other purpose. There are three properties of flash storage that are most important: latency, reliability and cost per TB. Certain use cases will require features such as replication, mirroring or snapshotting which are secondary but still essential in some environments.
I couldn't give a shit if an array dedups or not. I care about what it costs per TB. If the one that doesn't do dedup is cheaper, is more reliable and the latency is lower, then I'll pick that one.
I wouldn't choose an A or V 9000 unless I had to though. Like buying a 4K TV and using it with a VCR.