Automated-tiering storage solutions, put simply, have been developed to monitor application data workloads and automatically place the most frequently accessed data up into the fastest storage tier (usually SSD) while moving the less frequently used data down to slower performance HDDs.
The idea behind this is that the more expensive faster SSD technology is used mostly for the tasks in-hand and to increase availability of the SSDs while at the same time save money, the data that is not needed so often is stored on the slower less costly media of HDDs.
By implementing a tiering solution, organisations can benefit from only buying the amount of SSDs needed, instead of expensive over-provisioning.
In a virtualised environment where the IT infrastructure is more complex and there are dynamically changing workloads calling for 'hot' data to be present on the fastest storage medium at the instant it is needed, auto-tiering is seen as a must-have to realise the true efficiency benefits of virtualisation.
However, while these systems may seem to offer the best input/output storage performance and cost efficiencies, there are many pitfalls to be aware of.
Few auto-tiering products are perfect, with some being less automated than they appear and needing a lot of user-defined parameters or policies to be created before being operational.
It is not uncommon to see user manuals as thick as the old-fashioned telephone directories. Also auto-tiering systems tend to be aimed at the mid to high-end of the market so smaller businesses might not even get a look-in due to the initial outlay.
Although the biggest pitfall of all is that most automated tiering solutions do not operate in real-time but instead move a batch of data at a set time whether it is overnight or another off-peak time when it is least disruptive to the business.
The problem with this batch movement of data is that the benefits of having access to SSDs for the hottest applications is not realised sometimes until the following day which defeats the objective of having tiered storage in the first place.
The good news is that best practice has arrived in the form of smart auto-tiering. This evolutionary technology carries out data migration in real-time.
This means that bottlenecks at the storage level are avoided as tiering happens in a matter of seconds, making the storage infrastructure as versatile as the virtualised environments for which it is intended.
- Warren Reid is Director of Marketing at Dot Hill Systems and has responsibility for marketing both through and to Dot Hill's partners in the EMEA region. Prior to joining Dot Hill in January 2005, Warren was Technical Marketing Manager at Sun Microsystems.