The meaning of “serving” has some interesting connotations. I think of doing the bidding of others, of taking care to provide what others require and of impartially doing what is right. So it is with modern video servers. However, I think it is valuable to think of just how recent and shallow the roots really are, and how deeply they have modified the effective workflow in important parts of our industry.
Most servers, such as the Pinnacle MediaStream shown here, have gone through at least a couple of generations of hardware and many revisions of operating software.
My recollection is that Tektronix showed the first “video server” in public at NAB. It used JPEG compression, which at that point was used only for still-image compression. Concatenating successive still frames makes motion video and thus “motion JPEG.” The Profile video server was not so much aimed at a specific market, but it was shown as a development project with clearly some inkling of what potential they were releasing. But it stored too little material to be an effective spot server, had no real interface to the automation systems of the time and had lower perceived quality than the VTRs of the era. Concatenating multiple passes through the box was problematic unless the bit rate was set high. But the seed was planted. Within a year or so, it was being used as a cache on the output of robotic VTRs, enabling back-to-back playback all day long of 10-second segments. At sufficient bit rate it was arguably high enough in quality, and as a cache, large amounts of storage were not needed. Over the next few years, the most sincere form of flattery and imitation, brought many servers to market and radically changed the industry.
Today we find ourselves in a rather different time. Servers are a mature product. Most, such as the Profile, Leitch VR and Pinnacle Mediastream, have gone through at least a couple of generations of hardware and many revisions of operating software. Huge servers and small have been deployed. In some segments, the widespread usage of servers has displaced older technology virtually completely. One is clearly spot playback for television. Increasingly, servers are playing video in museums and even in point of presence displays in retail shopping environments.
Capacity has driven much of the change, as has the advent of low-cost servers without some high-end video application features. Without the current crop of low-cost devices, retail and other public space implementations would not have become as commonplace as they are. Behind all of this are some technical advances that have little to do with video and everything to do with computing.
Storage space of two or three hours was sufficient in early servers for only limited applications. Chassis were costly and adding large amounts of storage was cost-prohibitive. As disk-drive size crept up past 2GB and 4GB to today's 141GB and 181GB drives, and the cost for a drive fell from thousands of dollars to hundreds, the size of common configurations has dramatically increased. It is not unusual to see base configurations storing 60 to 70 hours, a 30-fold increase in a decade, at a real cost of approximately half. The drop in cost and increase in capacity together have removed barriers to storing long-form content on servers. Indeed, in the last four years, many large plants for DIRECTV, Turner Entertainment and others have been built around sophisticated architectures with all-server record and playout of programming. Together with archive management and media asset management (all from the IT domain), the rosy reality has made highly efficient plants with drastically different workflow possible.
The architecture of the storage was at one time a major issue to be considered. We are, however, reaching the point where the technology and topology of storage is less an issue than the total cost of ownership. For instance, in the Turner Entertainment plant, a complex environment stores copies on multiple servers, on redundant (RAID) storage arrays, and on DVD and data tape archives. The probability of loss of content is virtually non-existent. The content is moved from one medium to another under software control. It is no longer terribly important to know where the physical copy is, as it once was with videotape. It suffices to know that the media exists in one or more pools of storage (hopefully more than one). The method of storing can be debated by any two technologists available for the fight. NAS, fiber channel storage, and other options are available.
What matters is the performance of the total system (bandwidth demands and throughput available), MTBF and the standards supported. If you want to do H.264 or fractals, you will have a limited set of manufacturers to choose from. If you need garden variety MPEG-2, you have loads of options. If you need HD (SMPTE 292) I/O, there are few options, but if ASI I/O (or I or O) are satisfactory, the options open up.
Choosing a server comes down to verifying that the device meets the current and medium-term needs you can project, analyzing performance required (I/O, video-performance targets, concatenation effects, ability to transfer files directly via MXF), specifying the input and output connections (digital and/or analog, embedded audio, AES, number of channels), identifying the vendors you prefer to do business with, negotiating a price that is supportable for your application and verifying all of the above. All serious manufacturers will allow you to try out a box on your turf and generally are helpful in providing the data needed to analyze bandwidth capabilities of various configurations.
Lastly, you need to go back and look at special-purpose “TV needs,” such as automation interfaces supported, remote-control capabilities, stunt-mode features such as slow motion, availability of time delay or sports-replay software if it is applicable, and other unique needs of a television plant. Servers are not commodities when all factors are considered, but you will find a unique configuration that best suits your needs if you analyze the options carefully and thoroughly. Price is certainly not the only factor.
John Luff is senior vice president of business development for AZCAR.
Send questions and comments to: firstname.lastname@example.org