The decline of big iron: when using lower-cost tools, workflow and infrastructure are essential

Post, July, 2004 by Mike Hughes

Not long ago, the domain of high performance computing for the likes of weather analysis, oil and gas exploration, and seismic modeling was dominated by the likes of Cray and NEC, who built giant number crunchers designed for exactly these tasks. As with the dinosaur, they have now taken their place in history, replaced by the new, interlinked mega-clusters of cheap and small-form-factor PCs.

A similar trend is starting to take hold in the post production market as well. Call them what you will: the "hero" boxes, or more generically, big iron (dedicated black boxes or expensive hardware/software solutions), are being sold in smaller and smaller quantities. The move has nothing to do with a lack of admiration for these tools, but rather is a reaction to both technological advancements and business realities that have emerged over the past three to five years.

The single largest driving force in the migration to lower-cost solutions is the fact that desktop tools have matured to the point where they can play a significant and often complete role in the process. Developers of low-cost, hardware-independent Mac and PC-based editing and compositing solutions have come a very long way in the five years. Apple has led the pack with both its release of Final Cut Pro and a complementary purchase of Nothing Real and its Shake software. Adobe, Discreet, Pinnacle and now even Digital Domain (actually a spin-off called D2 Software) have added quality products to the mix. What we're seeing here has many parallels to the evolution that took place in the 3D market where the tools moved off the more expensive platforms as lower-priced, highly-appropriate hardware became available.

However, a move to lower-cost solutions is not without its perils and pitfalls. Price/performance gaps still remain, and higher adoption rates of HD and greater resolutions create imposing challenges in the desktop arena. More talent must be employed (and managed), albeit at a lower perseat cost. From a top-down perspective, there are two key components to a workable desktop-based approach: the pipeline and the infrastructure to back it up.

THE PIPELINE

The pipeline defines the flow of data through the facility--who touches what and to what end? How do matte painting, color timing, 3D, compositing, editing, etc, fit together? The pipeline must be rigid enough to ensure that procedures clearly define departmental responsibilities, yet flexible enough to account for the fact that every project is indeed different. It is the implementation of a functional and predictable pipeline that determines whether or not jobs can be predictably bid, and as such, a sustainable business built on the model.

Infrastructure is comprised of four main elements: storage, networking, input/output (I/O) and processing engines.

THE STORAGE

The storage component will define how the plethora of digital images, audio, textures, etc. are saved and hence from where they are to be accessed. It's a given that with distributed compute environments, a high-speed shared storage scheme is an absolute necessity. The two prevailing and often competing models are network attached storage (NAS) and storage area networks (SAN).

The network chosen is actually a close function of the storage scheme, with current approaches employing either IP-based protocols (e.g. Ethernet and ATM) or block level protocols (e.g. FibreChannel and iSCSI).

The I/O component defines how material enters the facility, to some degree how this same material flows through to those resources that require it and, finally, how results are delivered to the client. There are many, many different types of source material and almost as many ways to transfer them. From data tapes to removable hard drives to scanned textures to videotapes, the only thing they all have in common is that they must be digitized into the shared storage arena in a format that the applications in use can understand.

The processing engines encompass all CPUs, GPUs and custom hardware that are used in generating rendered results. From the users sitting behind their PCs and Macs, through to the rendering farms or clusters employed to offload the clerical from the creative, a properly allotted balance of such resources rounds out the facility infrastructural needs.

From an economic perspective, the move away from big iron is also a risk-reducing endeavor: Employing lower-cost hardware and a far more mercenary (in the best sense of the word) talent pool that often brings its own tools to the party and allows a facility to expand and contract organically as the level of required effort ebbs and flows with different projects.

As with life, nothing in the post business is purely black or white. The big iron that exists today is still in use, and a hybrid model is already in place in many locations whereby desktop environments feed data to a smaller number of high-end seats. Big iron numbers will continue to dwindle as support contracts come due and software versions fall too far behind. It's a cliche to say, but this will prove to be an evolutionary rather than a revolutionary occurrence.