Steve Atkinson

Storage Management Schemas for a Broadcast Environment

Abstract

As videotape gradually disappears from the broadcast plant, storage in a digital format becomes more important.  In most cases, this is a combination of spinning disk and removable media, either data tape or DVD.  With that comes the management of material as it moves from video server to nearline to archive and back again.  The class of software products which perform this function are generally called hierarchical storage management or HSM systems.  This is actually in misnomer in that HSM refers to a specific type of storage management rather than a product category.

Storage management really consists of four broad categories of products, each of which has advantages and disadvantages as it applies to on-air, production, and news environments.  Specifically, the categories are: shared filesystem, HSM systems, disk extenders, and data movers.

Shared filesystems came about in response to SAN systems being introduced into a heterogeneous computer environment.  In the IT world, they allow disparate operating systems to share files seamlessly.  This allows common storage to be used for multiple application and computer types.  This does not necessarily translate to a broadcast environment where video formats and real time processing require a more closed environment.  Shared filesystems are more reasonable to implement in a production environment where there is less of a real time requirement.

HSM systems are designed to migrate material between different levels of storage based on rules defined by the user.  These are often based on the fill level of the primary storage which initiates a move to secondary storage based on least used algorithms.  That is, the disk fills up to a certain level and when that threshold is crossed, the files which have not been accessed in a while are sent off to the next level of storage.  This is advantageous in that these rules execute automatically but they don’t necessarily have the rules needed for a broadcast operation.  For example, a broadcaster would not want the movie they ingested but have not played yet, moved off the on-air server ten minutes before air.

Disk extenders operate under the premise of making removable media look like a part of the spinning disk system.  The robotic system looks like just another drive letter.  This is implemented with a disk subsystem is the front end that provides the first level cache.  The disk extender keeps a “stub” of the file on disk and moves the majority of the file off to cheaper removable media.  The file looks like it is still on disk and when the application requests a file, the data blocks on removable media are restored to the cache and then transferred to the application.  This makes integration with existing applications very easy but it does have unintended consequences.  It is a two step process for moving for removable media to cache to application which can introduce a lot of non-determinism in when material will arrive at the on-air server, for example.

Finally, data movers operate as explicit migration tools rather than implicit migration tools as described above.  This means that data movers to move any files from one level to another until they are told to do so by a controlling application.  Controlling applications include automation, editing, and asset management.  This introduces a high level of determinism into the movement of files but also means that the process is less automated. 

This paper will discuss in detail the differences as well as the advantages and disadvantages of each approach.  There is no single approach which will work across the board for all broadcast applications and the implementation needs to be balanced based on a number of factors.

Biography

Steve Atkinson has been involved in high performance storage systems for the last 20 years with the last 15 years applying those storage systems to broadcast video.  Company experience includes Ampex, StorageTek and Avalon with his current position being VP of Sales, Americas for Software Generation, Ltd.  Steve has degrees in Journalism from Kansas State University and Computer Science from University of Texas - Dallas.