Starz Entertainment set out to replace its Hewlett-Packard Co. Enterprise Virtual Array 8000s last year when the aging system couldn't deliver the performance levels the global media company needed for its workflow system, particularly for its transcoding process that changes movies from one format to another.
The company chose DataDirect Networks (DDN) Inc.'s GridScaler scale-out network-attached storage (NAS), allowing it to condense nine storage racks into three while getting the benefits of a faster parallel file system for transcoding and encoding movies and original content.
Starz has one GridScaler per site, with one-and-a-half racks in each location. Starz's previous system included nine Hewlett-Packard (HP) Enterprise Virtual Array (EVA) racks that handled 750 terabytes (TB) of usable and 1.9 petabytes of raw capacity.
A Quantum StorNext file system was layered over the EVAs to present large LUNs to 40 HP blade servers for transcoding.
The older EVA system -- which has been retired by HP along with the rest of the EVA platform -- required more than 300 2 TB LUNs to be zoned across 40 transcoders and workstations. That made it easy for a LUN to get missed, which jeopardized the entire file system that was heavily dependent on the metadata server. Performance limitations caused transcode and encode processes to fail.
"If there was any dropped frame halfway into a movie, we would have to start over. We needed faster hardware," said Colin McGuire, Starz's director of infrastructure services. "One of our requirements was to get away from the StorNext file system. The combination of the aging EVA equipment and StorNext caused problems."
McGuire said the transcoding is done at the Englewood site and is a CPU-intensive process that demands high performance.
"With EVA, we would have to create job periods to copy larger files to move them to the disaster recovery [DR] site; that meant we did not have a full library because it took a long time to do that," he said. "We have to have a full library in the event of a disaster so we can continue to process media operations without interruption. At any given moment, we are transcoding up to 40 different formats."
With GPFS, the GridScaler systems write in parallel between the broadcasting and DR sites.
"We're dealing with writes as they happen rather than moving large files, which is a cumbersome process that has to be done at specific times," McGuire said. "GridScaler enabled us to create a totally different workflow [from] what we had in the past."