Since this is a blog related to oracle database storage optimization this post is an attempt to explain what we mean by tiered storage. You would think that everyone would have the same definition of tiered storage, but I've found that not to be the case. So, what is tiered storage and how does it relate to an Oracle database storage solution?
Tiered storage is the idea of creating different classes of storage with different performance, capacity and price points. The data lifecycle, in an Information Lifecycle Management (ILM) Architecture recognizes that data has different lifecycle characteristics centered around modification and access. Based on this knowledge a database storage architecture can be created to take advantage of that lifecycle if the right tools and infrastructure are available.
This is where Oracle Database 12c Automatic Data Optimization (ADO) and Heat Map come into play. These are the tools that are needed to provide a database centric ILM strategy. But wait a minute, storage vendors have already built tiering into their devices, why do we need anything in the database? The answer lies in the nature of how database data is accessed. Storage is aware of blocks, or files, but has no information about the meaning of the data in those blocks or files. The Oracle database on the other hand, understands that a table or partition is a logical entity, and with Heat Map can track that entity's usage characteristics. Why is that important? Again, the answer lies in what happens to all of the data associated with a segment at the storage level. Since storage doesn't understand database data, it can't know that the data it just moved to slower storage will be needed every 30 days for the month end closing cycle. Heat Map on the other hand will know this, and moving the segment can be avoided and not cause a performance impact. The converse is true as well, an ADO policy can be created that can identify segments that should be eligible to be moved to a different storage tier sooner than might otherwise occur if space pressure dictates. The policy can be created with the understanding of the data's lifecycle, something that a hardware device cannot do.
To conclude, I would like to stress that the storage tiers that I've been describing here are specifically for active Oracle database files. Since storage "tiering" seems to be an overloaded term, I want to differentiate the use of storage tiering in a database environment to help achieve Information Lifecycle Management as opposed to storage tiering that is referred to by storage vendors to differentiate between production storage versus development/test storage or even backup storage.