All-flash array storage, data reduction and the family SUV

13 February, 2017
James Surmacewicz
IBM

99,999 rolled forward to 100,000. After 5 years of driving my daughters to swim practice and dance class the family SUV has hit 100,000 miles with its fair share of door dings and soda spills.  It is time to start the adventure of buying a new SUV. In some ways it is exciting to start looking, but on the other hand going through the process is tiresome.  Interestingly enough, my SUV search happens to coincide with a parallel situation in the all-flash array (AFA) storage market.

Every new car has a window sticker.  It includes information about the car’s fuel economy. On the car I bought there was a large “21 MPG” sign in bold font. This is the estimated miles per gallon (MPG) performance of the car. But it is important to also notice what the small print says: “Combined city/hwy.” What’s interesting is that in the city we get 18 MPG and on the highway we get 25 MPG. Another way of looking at it would be this: if I were pulling a boat, I would get fewer MPG but driving down a mountain road my MPG would rise significantly.  The simple fact of the matter is MPG is dependent on the workload of the car.

Data reduction in all-flash arrays

This is where I see a connection to the AFA storage market.  Each vendor that produces an AFA has a “window sticker” on their data reduction rate (MPG). That rate is then used to calculate the effective capacity of the AFA. Data reduction is a group of technologies (pattern removal, compression, deduplication, thin provisioning) which reduce the actual amount of data you need to write to physical flash media.  This is important with AFAs because the less you store on flash media, the less expensive your storage system becomes.

AFA vendors also make their own claims on data reduction rates. But as an IT consumer, who should you believe? At least in the car world, there is a government agency which has certified testing results on MPG.  Adding to the complexity, data reduction technologies are deployed differently in competing all-flash arrays and the data reduction rate can even be measured differently.  So, as an IT consumer considering an all-flash array, what do you do?

Too often I see IT consumers fall into the data reduction trap. They get so focused on data reduction that they lose sight of other factors important to their overall decision.  Let me go to the car example again to prove my point.

The requirements for my SUV purchase were straightforward:

  1. 4WD (four-wheel drive)
  2. The ability to tow my boat
  3. 20 MPG
  4. Modern features (Bluetooth, apps, wifi)

With these requirements, I knew there would be some trade-offs. I knew the 4WD and towing requirements would affect my MPG negatively. I knew that I would have to pay extra for the technical features that I wanted.

Modernization allows you to tackle new workloads

It is important to remember why you are considering AFAs in the first place. Purchasing an AFA is about modernizing your data center.  Modernization allows you to optimize traditional workloads, which save resources (administration time, floor space, power, servers, SW licenses) that you can deploy to other projects such as analytics and hybrid cloud.

Selection tenants for all-flash arrays

Yes, data reduction is part of the discussion because it helps drive down the overall cost of the AFA solution, but I argue that you should look at data reduction as a single element of a larger solution and weigh it appropriately in your decision-making process. Know your workload and understand its data reduction potential. Be aware of trade-offs. I was willing to give up some MPG for 4WD and towing capability, and I was willing to pay a premium for modern features. This begs the question: which tenants should you consider when choosing an AFA?

Data Reduction Tenants V2

IBM can help. IBM has a wide range of all-flash array storage solutions for various workloads, capacities and price points.

And for those curious about my choice, I ended up buying a Jeep Grand Cherokee.

The post All-flash array storage, data reduction and the family SUV appeared first on IBM Systems Blog: In the Making.