Video servers - finding your best fit

It seems that the thing to talk about these days is video servers. The discussions take place in boardrooms, workshops, master control rooms and general managers' offices throughout the industry. Because there is a good chance that one of these discussions will be coming to an office near you soon, what do you look for when considering a video server system?

How do you wade through the near miasma of conflicting data, sales claims and the "best solution" of 25 different manufacturers? How do you hack and chop your way through the mountain ranges of buzz words - words like "rich media," "interoperability," "metadata" and "storage area networks?"

We're going to dispense with a nuts-and-bolts discussion of every single aspect of server systems. There isn't enough space in this entire magazine to fully cover the topic. We'll take it for granted that all video servers can ingest material, store it, retrieve it for playback and manipulate it in at least basic ways such as copying, transferring, etc.

We will focus on three main areas of technology here that have shown the most growth and development over the last year - data management, physical storage media and networking.

Data management Perhaps the single most important thing to consider in a server system is the data management system that it is connected to. Beyond storage size, beyond which MPEG flavor the system is using and beyond how fast material is being encoded, the ability to manage the data associated with that material is paramount. When your sales manager wants to browse a spot at her desk, she is going to need to be able to find it and retrieve it.

If you think that the people moving material in and out of the box will be the key to your data management system, you're only partly correct. Ask anyone who has had a basic nonlinear editing system in their plant for more than six months - they'll tell you that without strict, even draconian, rules and regulations, material can become hopelessly lost in no time.

A data management system has to be more than just an electronic filing cabinet. The great data management systems available today allow a vast array of relational database information to be kept on file. This is information that only has to be entered once. It's information that can be added to over time as a piece of material moves through the plant - information that can be searched, bundled, parsed, scoured and pulled out in an assortment of reports, lists and displays. Most importantly, it's information that never leaves the material it's associated with.

The best data management systems are based on standards. The Pro-MPEG Forum is working on the Material Exchange Format, or MXF. The Advanced Authoring Format Foundation is working on AAF. There are even signs that MXF and AAF may eventually coexist peacefully. Most of the major manufacturers are members of both organizations and will support both standards. This will be a world of media objects - a media object will end up being a video file, a video/audio file, an audio file, a still graphic, a set of animation cels, photographs, scanned images - anything that you might use in the "rich media" business we're all heading into.

If your server is going to be placed into or will migrate into a shared workgroup environment, a great data management system based on common, industry-shared interoperability standards will be critical. If you are looking for a small, stand-alone system that will be used by a limited number of people or run purely by an automation system, then maybe a more basic file system will do.

Physical storage media It wasn't very long ago that the only option you had when putting a server system together was to assemble a group of magnetic disks into a RAID array and start filling up your system with material. (And, boy, did it fill up quickly). When a "big" drive was between 4- and 9Gb, it was only the largest systems that could hold reasonable amounts of data. With magnetic disk prices running close to 50 cents per megabyte, even very large companies had a hard time assembling reasonable systems. This is where the "near-line" archive systems came in. Store a few hours of material on your very precious hard drive space, move it off onto relatively inexpensive data tape when it wasn't needed, then call it back from tape again when it was needed.

One of the most exciting developments in online mass storage this past year is built around optical media - previously relegated only to archive systems because of throughput speeds. Several manufacturers are now beginning to use new RAID systems built around fast DVD-RAM drives. A massive amount of relatively cheap, re-writable optical storage is now possible with read times as fast as 88Mb/s and write times as fast as 42Mb/s. With small optical RAID arrays clustered into storage area networks, we will very shortly get to the point of having many thousands of terabytes of material immediately available for a reasonable cost. Is this the end of the near-line tape archive? The signs are encouraging - systems available now are delivering optical storage at media prices in a range from about 1.2 cents to 2 cents per megabyte. This directly rivals today's average magnetic disk media, which is in the range of 0.6 to 1.25 cents per megabyte.

You can, for now, still get faster data retrieval speeds on some magnetic hard drives, so if your server application is using uncompressed data, or is expected to serve up material at rates faster than 88Mb/s, then you may still want to stay in the magnetic realm. If your need is for very large quantities of data available quickly without having to go to near-line storage, you may want to consider optical disk technology.

Networking Another area where exciting advances are being made now is in storage area networking. If you're not familiar with one, a storage area network (SAN) is a group of storage devices - hard drives, RAID arrays, optical devices, etc - arranged in a configuration that allows multiple workstations to use any of the material stored on any of them, usually simultaneously. The SAN has, until recently, been a sort of stumbling area for a lot of vendors. The main problem was one of resource sharing - because so many of the earlier server systems could only "talk" to one RAID at a time in a semi-dedicated fashion, a SAN was impractical for video and media.

A SAN is another place where standards are king. Storage area networking works well only if all of the workstations attached to the network can speak at least some of the same language that the SAN uses. Two things have made the SAN more viable in recent months. One is the fact that many of the vendors in the business have figured out how to separate their encoder/decoder boxes from their RAID boxes. The second is that a good-sized group of third-party vendors has stepped in to help out by marketing data engines, protocol translators and the like, which take some of the nuts-and-bolts level work out of actually making the various boxes talk to each other.

The fit is the key Selecting a server can be a daunting research task. The key, however, is in knowing what the system must do for your company. Once you have that nailed down, assembling the various technologies available into a system is almost easy. Like many things in broadcasting, it's a matter of finding the right tool for the job.