With the massive amounts of data that companies are storing, intelligent tiered storage isn't a luxury -- it's...
quickly becoming a necessity.
By Tony Asaro
I was sitting in a room of approximately 20 IT professionals from about 17 different companies and we were discussing the concept of intelligent tiered storage for SAN-based storage systems. Most of the people at this session worked for large companies with literally petabytes of capacity on the floor.
One of them was uncomfortable with the notion of the data storage system making its own decisions to move data to different tiers based on policies. He felt that some decisions should be made by humans. Two others in the group immediately disagreed. The "aha" moment came when one of them replied that with petabytes of data there was just no way they could make tiering decisions themselves. She said the storage system needed to be smart enough to move data based on metrics that would take people too long to analyze. She also pointed out that the problem will only get worse as their environment continues to grow.
Interestingly, this discussion was theoretical because at the moment none of these IT professionals have storage systems that can provide intelligent tiered storage at a granular enough level that is valuable to them. However, we've all been told by some of the leading storage vendors that this capability is coming in 2010.
Intelligent tiered storage is one of the most valuable capabilities on the horizon in the storage industry. Here's why:
- It's common for 60% to 80% of all data to become dormant 90 days after its creation. If you have 100 TB of data, then 60 TB to 80 TB of it is idly consuming lots of expensive primary storage each day.
- It's fairly universal that network storage systems have a capacity utilization of 50% or less.
- Storage capacity continues to grow for companies of all sizes and in all verticals. It was once unheard of for a small company to have terabytes of capacity and large companies to have petabytes, but it's now commonplace. This dynamic seems to be shaping up as a perpetual situation and, as such, tiering will become requisite.
- We know that tiering storage can save significant amounts of money. That's why we have different tiers to begin with. But tiering today is very rigid and requires innovation.
- Even though all of the above are true, very few companies have implemented intelligent tiered storage because it's not granular enough.
A number of storage systems allow you to set a policy to automatically move an entire volume, but that's pretty useless in most cases. If 10% of your volume is active and the other 90% is inactive, that volume is considered to be "active" and isn't moved to a lower tier. However, imagine if you could keep the active 10% on tier-1 storage and move the inactive 90% to a lower tier. The economics of granular tiering are essentially a "no-brainer."
Intelligent tiering is achieved by using technology that can "stretch" a volume across different tiers of storage. Tier 1 could be RAID 10 using solid-state or Fibre Channel drives. Tier 2 might be RAID 5 using SATA drives. The cost differential can be significant. For example, one of the IT professionals in our session actually converted all of their company's storage from RAID 1 to RAID 5 and saved more than $100 million dollars! Not everyone has petabytes of storage like they do, but this example illustrates the impact tiering can have and the same relative savings for companies of all sizes.
This technology works today and is provided by emerging SAN storage vendor Compellent Technologies Inc. with a feature called Data Progression. 3PAR Inc., EMC Corp. and Hitachi Data Systems do intelligent tiering at a volume level, and I believe they'll all add support for more granular levels of tiering this year.
The IT professionals in this particular session were cautiously optimistic that data storage vendors are heading in the right direction. There was consensus that granular intelligent tiered storage was going to have a major economic impact in their environments. However, they all felt that it wasn't enough just to move data efficiently between tiers; it's also essential to have good reporting and to ensure that all service levels are met without a hitch.
Innovation is still alive and well in the storage world. Over the last two decades there have been milestone features such as mirroring, replication, snapshots, logical pools, thin provisioning and data deduplication. I believe intelligent tiered storage will be added to this powerful list with the potential to enable major economic improvements in SAN storage.
BIO: Tony Asaro is a senior analyst and founder of Voices of IT.