No sooner than we start get to technology which allows us to practically get rid of tiers or at least the management of tiers by the introduction of automation to move data around arrays with minimal admin intervention; we start to bring a whole new bunch of storage silos.
There seems to be discussion about tiers again and how we categorise data; so we have SSD only arrays which are our performance tiers, we have SATA/NL-SAS which are capacity tiers and then we have tape/MAID etc which are our archive tiers. So here we go again, fragmentation of data across multiple tiers could well be with us again with-in the next half-dozen or so years.
And there will be hand-wringing again; ILM and HSM is too hard say the vendors, the returns are minimal. Once we see the mass introduction of SSD-only arrays or perhaps that should be an if; I think as storage consumers that we need to be putting pressure on the vendors to provide seamless ways to move data around between the arrays.
I wonder if almost inevitably, the standard operating model will be external virtualisation in front of silos? Or is it application level intelligence which does the movement? Or the file-system?
I am not against the introduction of SSD only arrays; actually I think starting from a different point as opposed to just putting SSDs into an existing array is not such a bad idea. Designing from the metal up may mean we see less compromises and more effective use of SSDs, yet it does bring challenges into the data centre.
At times, this all feels very circular.
