Managing Exponential Data Archives: Behind NCSA Blue Waters

Reading time for this article .
Subscribe to the Spectra Blog >

The following two words can often generate stress and concern for IT administrators: exponential data.  Exponential data means data sets with size and growth rates that exceed the budgets, storage footprint, and personnel resources needed to contain and manage them – especially over extended periods of time.  By 2020, storage growth will accelerate towards 40,000 exabytes and beyond, and avoiding exponential data will no longer be an option.

Creators and users of the largest data repositories on the planet are leading the response to exponential data.  The National Center for Supercomputing Applications (NCSA), home to Blue Waters, the fastest supercomputer on a university campus and one of the most powerful in the world, recently added 20 PB of storage to its existing Spectra Logic TFinity tape storage infrastructure.

Exponential data requires enormous scalability, performance, and reliability if it’s going to be harnessed productively.  The NCSA TFinity deployment was originally designed to house up to 500 PB of data while reading or writing data to 244 TS1140 Technology Tape drives at an aggregate rate of 61 gigabytes per second.  This combination of tape library automation and tape drive performance provides NCSA the storage scale, performance, and reliability it needs while delivering an economically rational solution that would be impossible to achieve with disk.

By adding 20 PB of storage to its Spectra TFinity libraries, NCSA is ensuring it can fully meet the exponential data requirements of its most demanding users.  For the scientists and researchers who depend upon the computing power and deep storage of the Blue Waters solution to model, design, and discover the most complex systems and theories known to mankind, NCSA and Spectra Logic have ensured that unchecked exponential data need not be cause for concern.  

For more information about NCSA’s Spectra Logic solution, visit: