Most people in the post industry would agree that a collaborative, file-based workflow with common central storage is the ultimate end game. Collaborative workflows have been proven by pioneering facilities to achieve huge gains in productivity, with commensurate business returns. But what is the best approach? Software solutions exist that allow most applications on just about any computer OS to fully use the shared central storage–but it is tough to guarantee real-time performance without overbuilding the system and blowing the next five years’ capital expenditure budget.

Another approach is a proprietary system that guarantees realtime performance–but only if you use a particular vendor’s workstations and applications, and only if you restrict your work to a limited number of SD and highly compressed HD seats. It’s unlikely that this same single-vendor approach can provide a solution for DI work, with its 2K and 4K formats reading and writing massive files on a daily basis.

Some facilities have created hybrid workflows that use shared storage for most tasks, but require direct attached storage (DAS) for those seats handling the most “complicated” work. This approach involves copying the massive files to and from the DAS, negating much of the advantage of shared storage.

Why is realtime performance so important? Almost all client-facing work, as well as most creative decision-making and quality control operations, require realtime performance. In the tape-based world, there was a simple formula: For each realtime stream, there was a dedicated VTR. If five streams were required, one bought or rented five VTRs. But collaborative, file-based workflows are not as empirical. The storage infrastructure that normally delivers five realtime streams may muster only one when the stars line up against you.

Why does this happen? IT hardware and software are designed to achieve maximum performance over an extended period of time, but with no guarantees of minimum performance on a continuous basis. This is often known as “best effort” and helps achieve impressive results for the money spent. This performance is ideal for email, databases, in fact, most tasks performed by computers today. It even works for significant parts of a facility’s workflow such as rendering and data movement. But it isn’t good enough for realtime film and video requirements where dropped frames and intermittent stutters are simply unacceptable.

The biggest problem in achieving consistent realtime performance is the comparatively slow pace of performance gains of disk-based storage. While computing power has grown three million times in the past 25 years, and network bandwidth by more than 1,000-fold, disk transfer bandwidth has grown by only seven times. Though it’s true that disk capacity has grown exponentially–you still effectively have a very small hose trying to fill a huge, ever-expanding tank. This problem is compounded in a shared storage environment with many individual applications simultaneously requiring reads and writes off the disk. Disk performance is even degraded by the simple desire to actually use all the capacity of the expensive storage that just busted your budget. The problem with disk drives and arrays is that once the storage fills to between 50 percent and 80 percent of capacity, performance rapidly falls off. That is because the disk becomes fragmented, and it becomes harder and harder to find contiguous areas on which to store the data.