The Storage Economist

Dynamic Tiering is a tactical (and yet strategic) method to reduce storage TCO. For so many years, we have built tiered islands with heterogeneous storage arrays, modular and enterprise mixed together. This segmentation of storage capacity has been exasperated with separate storage infrastructures for VTL, NAS, CAS, archive, etc.

Here is the basic premise of where some costs are hidden:

• Not all data is created equal. Not all data ages equally as well. Different tiers with different QoS levels present different costs (and total costs) for the enterprise
• Cost of data storage should not be the same. Many can argue that the cost of acquisition of a single tier is better than multiple tiers (due to aggressive discounting). This is very true but the multi year costs of these tiers do have TCO separation in terms of:
• Power, space and cooling
• Backup, local data protection and disaster protection requirements
• License costs for the software needed to protect the tier
• Management effort (troubleshooting, load balancing, etc) for the tier and QoS needed
• QoS of storage functionality is not the same between the tiers. As mentioned above backup and DR protection would vary significantly between high and low tiers
• The lifecycle changes of data over time changes the access rate and data value over time. Tiered storage enables ILM cost savings through non-disruptive demotion of files and volumes
• Cost of Copies is one of the biggest areas of cost impact with dynamic tiers. We see 10-15 copies of a database in an enterprise area to be used for test, dev, QA, regression testing, data marts etc. These copies do not have to stay in-tier. Copy-on-write volumes or snap volumes can and should be made to lower cost tiers. If you have a 400GB database with 10 copies, that can present a 4TB storage commitment in-tier. We may not be able to argue the need for 10 copies of the database (that is a business and application decision) but we can enable a lower cost offer for those copies
• Ability needed to promote and demote on the fly is an essential differentiator between dynamic tiering and static tiering. There should be little latency in promoting or demoting volumes based on time-of-month periods, pre-determined processing efforts etc.
• Dynamic tiering is also the premise for archive, being able to move stuff easily off and then easily back into the main production pool. IDC estimates that 31% of unstructured data can be considered for a deep archive.

Dynamic tiering can be archived within a frame (aka storage intermix) for smaller instances, or with virtualization to provide the tiers within externalized arrays.

In my Blog library (Bottom of this page) there is a MS excel worksheet that allows you to play with a 4 tier static model comparing it to a 4 tier dynamic model over time. For each of the 4 tiers you need to be able to estimate:

• Total usable capacity
• Growth rate per year
• Price per GB at acquisition
• Price erosion  year on year

To see the effectiveness of the model, you can define your current 1, 2, 3 or 4 tiered static storage capacity and then play games with the CAPEX difference over 4 years by have a re-distribution that better aligns to the age and use of the data. Here is the graphical view of the re-distributed capacity after 4 years for an environment with 140TB.

And with this re-distribution through dynamic tiers, the 4 year CAPEX savings is $658K while maintaining the projected storage appetite.

Have a try at playing with your results in a fully dynamic tiered approach.