Techno Musings

My flippant title is a bit intentional after all when tiering your data many people dock their data someplace to save it for a rainy day.  Sometimes that is indeed the case and perhaps files are placed in a long term tier for safe keeping. Typically we call this an active archive so that when that rainy day happens well there is a copy to recover from.  A lot of intellectual capital is spent on moving items from a more expensive tier to a less costly tier to maximize investment.  However the virtues of saving money by say moving from spinning disk to optical media or tape — cold media if you will — have well been evangelized to death.  What I want to talk about is that pool of high performance storage which is often thought of as costly, and indeed the capital costs with buying an expensive SSD/FLASH or FC/SAS storage media are high.  However why is high performance, high cost, high reliability storage used?  Well frequently it is to run the business.  So the comparisons one has to make is not about how much money one might save by deploying Tier 0 or Tier 1 class storage infrastructure, but how much money will be made by doing so or lost by not doing so.  For example, if your high performance storage platform enables you as a user or customer to perform 2 to 3 times the number of simulations than your nearest competitor then that is a valuable investment you can market this point to your customers, which should improve your ability to generate revenue.  While at the other end is say tape which is a cheap and as deep as you can get and it is really about cost savings and risk mitigation.

The really hard part of this is not about buying the right platform for your various tiers, that is really easy.  The really hard part is in determining which applications to put where.  Well it used to be that moving data between tiers was the hurdle and that storage architectures were largely islands.  Well that day is in its sunset.  Consolidation and non-disruptive data movement technologies are key to making a fluid infrastructure possible.  Essentially, I’m going to throw out there some thinking for how to initially place applications on a given file or block tier:

1. If an application does not contribute to the bottom-line with net positive cash flow it cannot be on Tier-1 or Tier-0, period.
2. All of the applications that are not positive cash flow generating should be on Tier-2, period.
3. If an application needs temporary access to a higher tier it should be moved to that tier based ahead of the temporary access and moved back down after the higher tier demand is required.
4. If an application moves from being negative cash flow generating to positive cash flow generating, sufficient Tier-0 or Tier-1 storage should be acquired and the application should be promoted.

Notice I did not supply any technology recommendations as different customers may want to describe tiers in their own way — BTW that is why both Hitachi Data Discovery Suite and Hitachi Tiered Storage Manager sport the ability for users to flexibly describe their tiers.  For instance a user may describe a tier as having certain latencies, with snapshots and remote replication as Tier-0, and another might describe it as Fibre Channel RAID-10 with synchronous remote replication.

Note that I’d be curious to see if anyone follows these rules or something similar, feedback and debate is most welcome.