Whether it’s to hold a table at a favourite restaurant, book a dream holiday or rent a car, making reservations usually pays off. But for storage and IT administrators feeling the squeeze of shrinking datacentre floor space and power constraints, the practice of departments “reserving” excess storage they may never actually need is becoming a major problem.
According to a new report by Forrester Research, mismanagement of storage capacity can drag down utilisation rates for block-based storage to as low as 40%. That problem can be exacerbated by factors compounding one another, including poor communication between the groups running storage and application owners, inefficient or overcompensating storage forecasting, and escalating volumes of data.
The study notes that for a storage department, the most visible measure of utilisation is the amount of allocated capacity compared with available capacity. Bad habits can form when owners frequently ask for more storage than they presently need so it can be safely tucked away for future usage. Unfortunately, they seldom realise that the capacity has already been paid for and resides on continuously empty spinning disks that feast on power and cooling resources for no good reason.
“One group makes a forecast and asks for way more than they need, and they usually get it,” says Andrew Reichman, an analyst at Forrester who authored the report. “Because storage has been very difficult to provision and fine-tune after the fact, IT environments have been forced to use a [paint] roller instead of a fine paintbrush to manage storage [allocation].”
While a capacity reclamation project is one option for recouping misappropriated storage, the process may cause more harm than good by disrupting applications. And it can sometimes prove difficult to retrieve excess space from owners hesitant to lose their existing storage cushion.
Storage analysts say that thin provisioning could potentially provide a remedy by helping to give application owners the illusion of reserving more storage than actually is given to them on the storage back end. By proactively reserving storage for individual applications only as needed, thin provisioning can have a dramatic effect on storage infrastructure efficiency and budget, says Steve Norall, an analyst at the Taneja Group.
“Thin provisioning is critical because it allows administrators to drive storage utilisation rates higher and thus get better hard-dollar savings,” says Norall. “Furthermore, it allows IT to defer purchases of new storage until they absolutely need it so they aren’t spending a lot more upfront on storage capacity that will essentially be allocated to an application but not used.
“This is actually a critical issue that we’re seeing in VMware environments, where people are doing server consolidation projects and need large upfront volumes but may only have 25% utilised [written to data].”
Still, as the Forrester report points out, thin provisioning is not without its share of risks. Factors such as potential application downtime, limitations with file storage, virtual server complexity and resource-hungry database deployments must all be taken into account before the technology is embraced, cautions Reichman.
Announcements in May by large storage hardware vendors such as EMC and Hitachi Data Systems that they are incorporating thin-provisioning capabilities into their respective product roadmaps are lending much-needed credence to the technology’s viability, storage analysts note.
Aside from Network Appliance’s FlexVols software, which concentrates on snapshots and file storage, native thin provisioning until recently has been typically delivered to the mid-market audience by smaller niche storage vendors such as Compellent, Lefthand Networks, 3PAR, Isilon Systems, EqualLogic and DataCore.