Issue link: https://digital.copcomm.com/i/851458
www.postmagazine.com 33 POST JULY 2017 4K HDR AND ATTACHING REAL PERFORMANCE METRICS TO 4K STORAGE CONFIGURATIONS BY DAVE FREDERICK SENIOR DIRECTOR, MEDIA & ENTERTAINMENT QUANTUM SAN JOSE, CA QUANTUM.COM ffordable 4K cameras and in- creasing demand for 4K/UHD TVs, computer displays and con- tent is driving the adoption of 4K video for film and video production. To create more immersive, engaging experiences and to capture the highest quality pixels for every shot, content creators are cap- turing and generating programming with greater resolution, frame rates, dynamic range, color depth and color gamut. They are shooting sporting events and action films in 4K and even 8K resolu- tions — at up to 120 frames per second (fps) — to present a tremendous level of detail, even in slow motion. These changes are transforming the video viewing experience. But to deliv- er this richer, more captivating content, many content creators need faster, more capable storage environments to capture, process, distribute and archive this high-quality video. Identifying just the right type of storage system is no simple task. The term "4K" covers a wide variety of resolutions, frame rates, compression schemes, bit depths and color character- istics. Understanding these differences and how they affect storage infrastruc- ture is crucial when the time comes to invest in a system intended to support a facility's present and future needs. Mastering in uncompressed, 4K for- mat consumes five to 10 times as much capacity per hour as compressed 4K. Preparing a full-length feature film in 4K can easily generate hundreds of tera- bytes of data. Depending on workflow requirements, multiple projects would require a storage environment that could scale from a few hundred terabytes to multiple petabytes. The performance requirements of 4K may seem to indicate that flash, with its tremendous performance, might be the best solution. But rigorous real-world testing of a variety of hard disk drive (HDD) and flash-based solid-state drive (SSD) systems offered a new — and val- idated — perspective on the ideal use of HDD and SDD to support 4K workflow. Testing completed by Quantum in 2017 covered more than 14 storage configurations and their performance in supporting a 4K environment. They were benchmarked using three com- pressed 4K formats, ranging from 111 to 249 Mbps, and three uncompressed media formats, ranging from 807 to 1990 Mbps. Client operating systems were also tested with Windows and Mac OS-based systems performing about 15 percent slower than Linux systems, due largely to the number and type of I/O messages for each system. All client workstations were connected to the storage arrays using 16 GB/s Fiber Channel connectivity. Autodesk Flame and Blackmagic Design's DaVinci Resolve were used as a real-world benchmark to calibrate stream counts against VidIO, an automated test application. Each storage array was filled to 85 percent with test material sized to match high-quality RGB data and not the less demanding YUV data that is sometimes used for benchmarking. For uncompressed media, VidIOcreated frames to match published resolutions and data rates; for compressed media, individual files that mirrored published codec bitrates were used. Considering the dramatically high- er resolution of 4K media, it would be reasonable to assume that greater 4K stream data rates would have the highest impact on storage arrays. However, the test results revealed that the higher num- ber of compressed 4K stream counts impacted storage array performance just as much as stream data rates. In the case of HDDs, supporting concurrent compressed streams caus- es drive heads to race to read multiple streams, which increases latency and takes a toll on performance. Thus, while the cumulative bandwidth of individual compressed streams may be much less than the theoretical maximum perfor- mance of the array's controller, the laten- cy of reading high numbers of streams can quickly overwhelm HDD-based arrays. As a result, arrays with faster, 10K RPM 2.5-inch HDDs are better suited for higher compressed stream counts than ones with slower 7200 RPM 3.5-inch HDDs. In comparison, these lower-cost, higher-capacity disks could produce the same number of uncompressed streams as the higher performing 2.5-inch drives. SSDs proved ideal for very high-com- pressed 4K stream counts as long as low capacity was tolerable to the use case. Compressed streams can leverage the random I/O performance and lower latency of flash better than uncom- pressed workflows. For uncompressed work, you will find a spinning disk a much better value, providing similar stream counts to Flash, but at much lower prices and greater capacity. The value of disk capacity is particularly im- portant in uncompressed workflows, as this is where file sizes will be the largest. Ultimately, in terms of price, perfor- mance and capacity, the sweet spot will differ for every organization and 4K use case. Now, however, media organiza- tions can use real-world performance and scalability guidelines to match their needs against various vendors' solutions. Those interested in reviewing the results for themselves, can download the complete white paper at: http://landing. quantum.com/NA-4K_Ref_Architecture_ LandingPage.html. A