The magnetic attraction of RAID storage

The magnetic attraction of RAID storage

Optical discs have long been the simple choice for large imaging projects, but magnetic disks now have their own advantages, says Peter Webb.

Optical disc technology has long been associated with imaging, and optical disc jukeboxes have been the "enabling technology" of the industry. For many of the largest applications, they have been (and remain) the only practical way of storing the large amounts of static data generated by scanning millions of pages of data.

However, it is also clear that the use of optical discs and jukeboxes within imaging systems is declining, and the use of magnetic disc (particularly RAID) for on-line storage is increasing. The benefits of optical discs in imaging are generally considered to be cost per byte of storage, incorruptibility (data cannot be changed), and longevity (the media lasts for many years). All of these are under threat from both technological development and the changing nature of imaging applications.


This remains one of optical's great strengths. While optical drives are considerably more expensive than magnetic drives of the same capacity, optical discs can be mounted in jukeboxes. That means a single optical drive (or in practice two to four optical drives) in a jukebox can service dozens or hundreds of discs. Doubling the capacity of a jukebox is simply a matter of doubling the number of slots ("crypts") available for the discs themselves. If you have very large amounts of information, optical disc jukeboxes are very cost effective.


In theory, information held on optical disc cannot be changed. In practice, optical discs usually rely on databases held on magnetic disc, and these must also be secured. The unchangeable nature of optical is generally considered an advantage for document imaging, as the images cannot be altered. However, the imaging industry is currently converging with the DMS (document management system) industry. DMS software is far more concerned with changeable objects such as word processing files and spreadsheets, and accordingly unchangeable media such as optical discs are a problem rather than a neat solution. As imaging and DMSs have converged, optical has become less attractive.


According to accelerated aging tests, most optical discs will last for between 30 and 100 years. In comparison, magnetic media (cartridge tapes) require periodic re-tensioning (rewinding) or refreshing every few years. Of course, just because the media lasts 30 to 100 years doesn't mean that you can recover the information after that time - you still have to be able to find a system to read it on. There are no standards for how data is organised on optical media, and accordingly every disc is effectively in a proprietary format. In general, you can't take an optical disc off one image system and read it on another. This stuff is hard to do now. Finding a system to read a proprietary format on a niche media that hasn't been used for 30 to 100 years is going to be a challenge.


Optical jukeboxes are slow - typically 10 to 20 seconds to load a disc and retrieve an image. Magnetic discs (eg RAID arrays) should allow the image system to operate at the same speed as any other client/server application - perhaps one or two seconds on a fast imaging system. Vendors offering optical disc like to point out that 10 or 20 seconds is much faster than a manual retrieval. However, few of them like to bicycle to work on the basis that it is "quicker than walking". Fast is good, and jukeboxes are definitely not fast.


This is a major strength of optical. Large magnetic archives are based upon use of RAID technology. While different types of RAID use different approaches, the usual concept is that each file is spread over several physical discs, with error recovery protecting against a single disc failure. However, in the rarer situation of losing two or more discs, the whole RAID array is unavailable.

This is attractive for "normal" applications, where the data is needed on an "all or nothing" basis. In imaging, however, losing a single disc might only affect a small subset of your files. Most imaging systems allow direct one-for-one backups, where a backup disc can be directly loaded into a jukebox without any explicit software recovery process. Similarly, if the jukebox itself fails, some systems allow a dedicated operator to manually load discs until the system is repaired. This wealth of recovery options means that optical systems can potentially have better availability than magnetic systems, which in some applications is extremely important.

Peter Webb is a principal consultant with information management consultancy Opticon.He can be contacted on 0413 737 509 or at