The Storage Economist

Hu Yoshida seemed to hit a nerve a few weeks ago with some blog comments around tiering, the economics and practicality of in-box tiering (aka intermix) and the out-of-box tiering through external arrays/virtualization. Depending on how you architect and deploy tiering and the static or dynamic nature of the tiers, your economics or unit costs will be different.

First, I will not spend time detailing many of the econ impact areas of tiered storage. You can read an old white paper of mine on this topic, or some other papers from other sources to get up-to-speed. This one from Gartner is a keeper…

Perhaps there are some basic questions and assumptions that need to be outlined

1.    So storage tiering is good? Yes
2.    It can save you future costs? Yes
3.    It is free to implement? No
4.    It can save you lost of money right now? Not necessarily
5.    It is easier to manage? That depends

Item #1. Storage tiering is good. Putting data in the right tier with the right costs, SLA, OLA is key to holding down OPEX costs. All these papers will convince you beyond doubt of its goodness.

#2. Will it save you future costs? When you tier the storage you also tier the cost of growth, assuming that your data and storage growth requirements will behave in the same tiered manner. Purchasing more Tier 3 or Tier 2 over time will be less than purchasing all Tier 1.

The graph below is from a client, and you can see 3 years of total capacity growth, but with different distributions and growth within the tiers. You can guess that if the growth in 2008 and 2009 had been in tier 1 only, their CAPEX spend would have been much higher. This re-distribution and investment (less in tier 1 and more in lower tiers) saved them millions in CAPEX during this time.

Power, cooling and floorspace as a function of Cost/TB/year will be lower with multi-tiers.

#3. Is tiering free? Probably not. Software to manage, move and redistribute (automatically) tends to cost money. If you do not own a storage service catalog (to define and defend your tiers), then you have to take the time to put this in place. Processes and business rules will be needed to define data promotion and demotion.

#4 – Tiering tends to save longer-term, cost of growth or CAPEX costs. I have seen cases when a customer has 1 monolithic tier only, and rather than purchase more Tier 1 they make the investments (see #3 above) and purchase a lower tier storage pool. The CAPEX is lower for that one time purchase, but those savings are often off-set by the investments.

Often I find situations where the cost of copies is very high. Customers need copies of data and databases. Some need 2-3 copies, others need 10-15. The cost of copies is high when the copies are in the original tier where the primary data is located. Implementing tiered storage can allow you to keep all those copies, but in a lower-cost container. In these situations, tier 1 disk space can be reclaimed and replaced with lower cost storage, and that just might be enough to defer tier 1 disk where you can claim a more immediate cost saving impact.

#5 – It is easier to manage. I have a Storage Architect /colleague at CSC in Manchester England, and we have debated this topic many times over the years. One argument that managing, sorting, promoting, demoting and setting rules for 3-4 things is more complex and harder than 1 thing to control. There is, in fact, a point where you can have too many tiers, and you will notice diminishing returns.

The key to finding your economic sweet spot should include several assessments:

• Nature and aging of data
• Tendencies towards copies, long term archive, promotion/demotion of data
• Growth rate of all types of data
• Maturity of the storage management team, and alignment to business requirements
• Existence and use of storage services catalog, storage architectures
• Ability to move and adjust storage cost/performance parameters non-disruptively
• Justifying tiered storage with the kinds of money that are important to your organization