When it comes to tiering, it’s worth listening to what storage vendors have to say. There are actual benefits, and several approaches with real competitive differentiation.
Performance and cost are the big benefits of storage tiering. The goals are to balance the workload across the different storage elements and maximize the value returned from the investment.
Using SSD in a tiering approach provides improvements in latency and transfer rate. Intelligent tiering can maximize SSD usage as a resource, which in turn can reduce the number (and cost) of SSDs in a system.
If you’re considering adding tiers or changing your tiering strategy, you might first want to look at the available methods.
Storage systems have tiers usually consisting of several types of hard disk drives (HDDs) and perhaps SSDs. There may be two types of HDDs – a high RPM drive with less capacity for higher performance and a lower RPM HDD with higher capacity. Customers seem to be moving towards SSDs for performance and larger capacity HDDs as the tiers of choice. The movement of data between tiers is based on analysis of access patterns to data.
Most of the tiering systems being offered by vendors now have sub-LUN movement of the data (early implementations only moved entire volumes/LUNs). Key choices to make around the implementations deal with the “real-time” analysis and the automation around the movement of the data. Another difference for consideration is the potential contention created by the movement using system resources that are also needed to satisfy host requests.
Another tiering method is through intelligent placement of sub-LUN segments of data along with caching tiers. This approach uses the analysis of patterns of data access to determine where to place the data, and cache tiers to control performance. The goal with this approach is to provide the performance based on data access characteristics without introducing additional resource consumption for movement.
There are also hybrid approaches where intelligent placement and caching are combined with limited background movement. These approaches try to strike a balance where the intelligence of the tiering controls the resource usage while balancing the potential performance gains.