Managing capacity planning with thin provisioning

Infinity Pharmaceuticals uses thin provisioning in 3PAR Inc.'s InServ S800 system to manage capacity planning for scientist's hard-to-predict data storage needs.

While thin provisioning is used at many sites for utilization improvements, it has another important role. In environments where capacity needs vary widely, the technology can allow administrators to provision large chunks of capacity for individual projects with unknown storage needs without removing that storage capacity completely from use.

Peter Herrin is IT manager at Cambridge, Mass.-based Infinity Pharmaceuticals Inc., which develops medicines for cancer treatment and related conditions. He uses the thin provisioning in his 3PAR Inc. InServ S800 storage system more to manage storage capacity loads than anything else.

Herrin started with Infinity Pharmaceuticals in the summer of 2002, just about when the company purchased the 3PAR system and 5 TB of capacity. Herrin has since increased the system's capacity to 24 TB.

Before Infinity Pharmaceuticals acquired 3PAR's thin provisioning technology, provisioning storage capacity for the company's drug development projects was a nightmare. Herrin said scientists told him that each development project would require between 100 GB and 2 TB of storage capacity within 12 months. "That really doesn't help me," Herrin said.

He would initially provision 50% of a project's anticipated capacity, but would have to frequently shut down servers and migrate them to arrays with more capacity when they grew beyond their allotted space.

With thin provisioning, Herrin can provision all of the scientists' requested capacity knowing that the system will only allocate capacity when needed. "The thin provisioning in our environment has really helped because of the uncertainty in the science," Herrin said. "When we start a project, it's not quite clear how much storage is going to be needed for the project. How many test runs the scientists do dictates how much data is actually stored."

Raw data from the test runs comes directly into the storage system. It's then parsed into numerous Oracle databases. The storage system holds the original data from the test runs and copies the data to databases. When Herrin first started at Infinity Pharmaceuticals, he was provisioning storage for new projects approximately every two months, and the capacity requirements fluctuated wildly. The company also does a lot of database development, which requires a lot of storage capacity to migrate, change and upgrade the databases. In addition, the storage system keeps all of the company's typical operational data.

Herrin said the thin provisioning technology has supported every application it's faced without issue. "We haven't seen any application not respond to it," he said. "No hosts or OSes have had any problems with it. It functions flawlessly."

The only problem he's seen is when his scientists fill the databases faster than the system can allocate thin provisioned space. Herrin said the 3PAR system allocates space to the databases based on their usage history. He deals with the problem by limiting how much each database can grow in a specified amount of time, which slows down the data entering the system.

The only other hitch in his experience using thin provisioning is with the licensing. 3PAR licenses its thin provisioning technology by how much capacity it manages. So realistic capacity planning is key. "When designing your system with thin provisioning, keep the licensing in mind," he said. "If you know a volume is going to fill up within a few months, then it's probably not worth thin provisioning it. If you're more uncertain about the amount of space required, then it's definitely worth it."

This was last published in October 2009

Dig Deeper on Storage optimization

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.