There are several factors in selecting a hard-disk drive storage subsystem. For broadcasters and other content-origination suppliers, the storage choices are generally dictated by the broadcast media server equipment manufacturer, who will closely match the various storage, switching and encoding/decoding components in order to optimize performance.
The decisions on how to package a storage system for a broadcast server is often influenced by both product marketing and engineering performance. The mystical concept, which infers that "bigger drives are better" because "larger numbers must mean better," is driven in part by myth and fact.
(click thumbnail)As the internal disks spin faster, the time it takes to retrieve data (latency) is reduced.Other influences on a storage solution that are sometimes misunderstood include factors in the structure of the drive array and in the individual hard drive component itself. While logic in the bigger-is-better scenario seems to follow Moore's Law, the real impact may be obscured somewhat by misinterpretation.
Issues such as size and speed for discrete drives play well into desktop PCs, single-ended streamers for media and unprotected or limited bandwidth requirements for media storage systems. Without doubt, as compression gets better, less space will be required to store similar quantities of like-quality moving media. Using H.264 or Windows Media 9 (and VC-1), high-resolution images can be stored in 30 to 50 percent of the space required for the same images compressed with MPEG-2.
Furthermore, the platforms that display high-density images may not require the performance that current high-end MPEG-2 systems need. Future systems may indeed allow single-spindle, high-capacity drives to play a much larger role in the storing, buffering and real-time playout of moving images.
These new encoding advances result in more media being captured in a similar storage footprint. Besides having the same media consume fewer bits per unit, the need for higher throughput (disk bandwidth) is also reduced--allowing more program streams to play from the same storage subsystem.
COST & CAPACITY
While alternative compression formats for moving media are already in the marketplace, the issues surrounding current implementation and expansion of storage capacities on media servers continue. Until magnetic storage reaches the point of aerial density saturation, drive manufacturers will make physical systems to address the demand for growth.
Some of the pertinent points in the size-versus-speed categories will be explored in the balance of this installment, starting first with a comparative look into recent advances in disk drive technologies.
Capacity-versus-speed discussions become intermingled when the cost/capacity question is raised. In mid-2004, the 36 GB (10,000 RPM) drive was widely selected, based on its price out of the box; at the same time, the alternative 146 GB (15,000 RPM) and the 300 GB (10,000 RPM) drives provided solutions at a highest initial cost but lowest cost per gigabyte. These are the issues that have to be balanced with regard to disk performance.
Since 1987, when disk drives held around 20 MB, mid-2004 capacity has increased some 15,000 times. When compared to the other factors in computer technology, the hard-disk drive is way behind in performance growth. In the same period CPU performance and memory size have both grown by some 2 million times; with memory performance increasing approximately 50,000 times. This means that disk drives are expected to perform under much greater stress than other server components in a system.
(click thumbnail)Data access time affected by rotational latency and actuator seek time.The higher drive speeds (15,000 RPM) have reduced the impact of the lagging disk drive system performance, because as the internal disks spin faster, the time it takes to retrieve data (latency) is reduced. Both actuator seek time and disc-rotational latency are actually improved by about 30 percent when employing 15,000 RPM drives, according to a leading drive manufacturer.
To address the impact of data access from slower drives, storage system manufacturers have employed RAID 5 architectures, which then spread the data across multiple drives. This comes with a cost but also brings some gains in overall bandwidth and protection during drive failure Whether the drives spin at 10,000 or 15,000 RPM, overall system performance is a balancing act between using more "slower" drives and fewer "faster" drives for a given performance expectation.
THE TRADE-OFF
With fewer large-capacity drives, there is less supporting infrastructure and space, but lower maintenance and management costs, and higher I/O per second per unit (IOPS/U).
Over the past two decades, we've grown to understand the cost of videotape transport upkeep and tape pricing, but the cost to maintain a group of media servers may not yet be as clearly understood. As the demand for storage increases, so will the cost to maintain larger servers--something probably not projected in the average station because of a lack of operational experience compared with the use of videotape.
Another factor with faster drives is that when a failure occurs, the time to rebuild the RAID decreases. Furthermore, there may be lower costs from the reduced power and cooling requirements, and with noise or vibration also reduced, the work environment improves and longevity of the drives increases. All these factors begin to interplay in a deeper perspective once the total cost of ownership (TCO) is considered in a system design.
WORK & TIME
In retrospect, the analogy that "work tends to expand so as to fill the time available" probably will remain applicable to the storage of moving media for quite some time. Given the choices and changes in storage architectures, it becomes all the more important to understand as many of the issues as possible, as early as possible. While equating IT models never completely parallels moving-image models, understanding the impact of those issues is essential to a long-term storage solution regardless of the enterprise.
