I recently attended the Fujifilm annual Global IT Summit in Houston featuring CIO and IT executives from around the world. At the summit the subject of using tape in production came up. While we spend most of the time using tape for backup and deep archive there may be some workloads that could actually leverage tape in production.
As I discussed in my first column reviewing the Global IT Summit tape, from a technology standpoint, has many of the right pieces in place to take a more important role in production applications. But how do you decide which workloads should move into production and how do you develop an architecture that will support tape well in production?
Compatible Workloads Production Tape
The key is to find the right workload. As I have written about in the past that means changing the way we think about when data is moved between storage tiers. Instead of looking only at the age of data we need to look at the predictability of its access. The more predictable the access pattern the more likely it is that tape can move from an archive tier to a production tier.
For example, in a media and entertainment organization the work is often done in a sequential process. In many cases users (and applications) know which set of data are going to be used and when. These data sets can be pre-positioned based on that workflow. When a project is done or has reached an operational pause, they can be moved to tape. When a project becomes ready to get started again, it can be moved from tape up to hard disk or even flash based storage.
Another candidate is any data set that can be streamed. In these examples there is often a large data file that’s sent to a user for immediate consumption, with audio and video files being the simple examples. The first part of these files can be kept on disk, so that the application can start processing immediately, but the rest could stream from tape. Tape, once it begins streaming, can sustain the ingestion rate of just about any receiving application or client.
In both of these tape-as-production use cases, the delay for a recovery from tape is unacceptable. Unlike archives, where data is often buffered by disk and users know an occasional restore may come from tape, production data must be available. Consequently, these data sets must either be ‘pre-positionable’ or ‘streamable’ so that by the time data is coming from tape the user is already starting to work on the data.
Tape is a real production storage alternative. We’ve shown that the technology is real and the workloads that can leverage tape are plentiful. And, the economics certainly make sense. In both the examples listed above you can reduce your high performance storage cost by 80% or more. The next step is to actually design the architecture, something we will cover in our next column.