What this means is that a quick look at the storage system would convey that it is full and in need of an upgrade, however, with a closer look, not all of the allocated storage is being used. Likewise, even though the database areas show that they are using about 90% of their allocated storage, if you were to go in and look at how full the tables are, there are areas of under-utilization, perhaps for growth, performance, or other operational concerns. The net result is that there is still plenty of data storage space available, and if that unused physical storage capacity could be made available to others, there would be an increase in effective capacity.
Thin provisioning works to overbook a reservation for storage capacity, leveraging the fact that there is allocated and unused storage capacity available. Building on the above example, if thin provisioning is used, 3 TB of storage is still allocated to home directors and file shares, 3 TB for email and 3 TB for databases. However, leveraging the unused physical capacity, there is more than 5 TB of unused storage space that could be allocated for a different application use. This extra storage space is often used for overbooking, which is something many companies that rely on reservations for space-based services such as airlines.
The thing about overbooking is that more often than not, you hear about it when things go wrong, which may seem like all the time, yet in reality this does not occur as often as perceived. Airlines, for example, have information and usage profiles to determine how much and what flights to overbook, and they get it right more often than perceived. With thin provisioning, when the knowledge and history of information is allocated along with usage patterns, intelligent decisions can be made to more effectively utilize available physical storage capacity.
When thin provisioning is done correctly, storage capacity is used to its fullest while simplifying routine management functions. The caveat, like other forms of overbooking, is that tools are needed to remediate when there is a conflict to resolve for double-booked resources. Another caveat is thin provisioning primarily addresses data storage capacity and not performance, so keep that in perspective when looking at different vendor implementations. The solution is to have tools that provide insight into usage along with performance activity patterns. These tools should also be able to take proactive action by adding extra physical capacity when and where needed.
So next time you are at the airport and complain about your seat being changed for a flight due to an aircraft swap, your trip may have been protected by the airline spotting a trend, having a larger aircraft available and doing some just-in-time capacity changes due to thin provisioning.
This was first published in December 2009