Equipment racks at WCPO-TV house the Omneon Spectrum media server system.
My dictionary tells me that a server is “Computer science: A file server.” That's not very illuminating, unless you think about it a bit deeper. Video servers erupted onto the media landscape as computers whose function was to playback video on command (not on demand — that's a different context for another time). They were simple beasts, with puny disk drives we once thought were huge. They produced acceptable pictures using motion JPEG at something like a whopping 24Mb/s. Yikes! A 2GB disk and 24Mb/s? Not only that, they were pretty simple devices responding to commands intended for VTRs. The dialect included no selection of clip name, nor trick mode playback initially.
As in all things, the bleeding edge of technology rapidly gives way to real products when the market thinks the engineers and marketers were onto something important. Ghandi said, “Imitation is the sincerest flattery.” We certainly had lots of flattery, and quickly. Servers sprang up faster than spring flowers. Many used the same components and simply different hosts. Some tweaked the codecs to achieve better performance. Within a few years, MPEG-2 codecs allowed significant improvements in storage density, and over time the standard record rate dropped by 50 percent, and more, for good quality.
But in the age of TiVo, what is a server for professional use? After all, I can record a couple hundred hours of HD and SD content and browse the metadata using a remote from my couch. And like it or not, the line between a “video server” and home recording technology is exceedingly thin. What separates these tools, and where are we going?
As in all things, it is not simple. Though consumer electronics are remarkably reliable and sophisticated, the line between these different genres is pretty long. Professional servers in general have intelligent remote control capability these days, and they usually respond to complex command sets that allow an automation system (newsroom or play to air automation) to send the name of a clip and a command to cue, park and play on command, with frame accuracy. Consumer boxes don't need frame-accurate results. The operative rule might be, “Just do it before I get perturbed and hit play a second time.”
Quality is important in both, but consumer boxes are not perceived as bastions of quality maximization. When recording is integrated into a receiver for DBS services, the quality of the recording has precisely nothing to do with the local codec because the bit stream is simply parked on a disk and later played back verbatim. When a local broadcast signal is recorded, the codec need only exceed reasonable consumer standards long accepted, hopefully of quality higher than VHS. The consumer box must function reliably. The professional box must function precisely on cue, meet rigid interconnection signal standards, have high quality and be able to be upgraded.
Storage density is less important in the professional box than total quantity of storage. Though early servers held a few hours, today servers routinely have 200 hours or more and are much smaller in size than units of a few years ago, especially if storage time is equivalent.
Another distinguishing factor is the number of I/O channels. Simple servers and consumer boxes are generally one or two inputs and perhaps as many as two outputs. Commercial server solutions must be adaptable to tens of outputs for many applications in a seamless whole that shares a common pool of storage, often in the hundreds or thousands of hours.
Consumer boxes don't need to interface to archives, nearline storage subsystems, asset management software, newsroom automation systems and the like. Though likely to change in the near future, consumer boxes do not need to take files in from other servers (compute and video servers) and handle playback. Professional applications must deal with commercial and content delivery systems that drop content on intermediate storage locally, including the movement of that content to playout systems.
And so it seems there are plenty of defining differences. Upon closer examination, however, we find the types of professional video servers growing and the differences becoming more of a continuum with consumer applications on one end, and large-scale playout systems on the other. In between, we find systems designed for point of sale applications, edge servers for broadcast networks, production caches in video switchers, and quite literally dozens of other applications that did not exist 15 years ago, or in many cases even five years ago.
The low end of the professional market is populated with many systems costing less than $15,000. They perform the function of VTR replacement admirably. Some systems are used for playout servers in small stations with “challenged” budgets. Some of these servers use commercially available codecs, which can be bought for less than $1000, and combine them with industrial computers and unmodified off-the-shelf operating systems. They fit a niche in the market that is important. Many production control rooms include such systems for playback of repeating backgrounds, bumper clips and transitional elements.
As you move up the price range, you find a crop of general-purpose servers used in broadcast and ancillary uses. The available storage options are likely to include networked storage systems (NAS, for instance). They range of options for expansion of I/O, storage, and archive and network connections is enlarged from the low-end systems, as is the quality of the I/O in general due to more engineering dollars available to optimize performance. Reliability is arguably better as you move into this market segment ($15,000 to $45,000).
At the high end of the market, one finds rugged, complex systems with high throughput to allow high bit-rate operations, including HDTV record and playback. Until recently, only a couple of manufacturers had HD codecs internally (SMPTE 292M). Others relied upon “best in breed” solutions from codec manufacturers and used ASI as the I/O. This year, one manufacturer introduced a line of servers that uses fast processors to allow software codecs to handle the input side, with more traditional hardware decoders. Bandwidth on the backplane is increasingly important in this market segment because the number of I/O channels may require more than 500Mb to permit simultaneous record and playback of multiple channels.
Where are we headed? The upper end of the market will likely remain complex and relatively expensive. The bottom end of the market will improve with more reliability and better performance as H.264 codecs become commercially available at reasonable prices in the future. And the consumer market will grow in capability to look even more like the big brothers we traditionally call video servers.
John Luff is senior vice president of business development at AZCAR.
Send questions and comments to:firstname.lastname@example.org