EBU • UER - Technical Review

No. 294 (April 2003)

The death of tape?

Regular readers of EBU Technical Review will probably know that I collect data on items related to broadcasting and electronics. For example, Fig. 1 is the latest version of a graph showing the price trends of hard disks and random-access memory (RAM). It shows that dramatic reductions in the price per MB have been sustained over many years.

Figure 1: Price per MB for hard disks and for RAM

Fig. 1 also confirms that the price per MB is falling more slowly for RAM than for hard disks: in 1990, RAM was approximately 10 times more expensive than hard disk storage but, by the year 2000, this ratio had increased to about 100. In the mid-1990s, it was often said that these downward trends in price per MB could not be sustained in the long term. Although Fig.1 indicates that such fears were misplaced, some caution is necessary when extrapolating these curves into the future. As with shares on stock markets, it is important to recognise that "past performance is no guarantee of future performance"!

Figure 2: Price of storage media as a function of storage capacity (December 2002)

Fig. 2 summarises another set of statistics covering various types of storage, such as RAM, hard disks, optical disks and magnetic tape. In this particular case, the price of storage has been plotted against its size: for example, a CD-RW (rewritable CD) offering 640 MB of storage costs about US $1. As there are various types of RAM, Fig. 2 shows the data as a series of points connected by straight lines. Note that the dotted blue diagonal lines represent the price per unit of storage (e.g. per MB or per GB).

Having compiled this graph, I was amazed to see that there was hardly any difference in the price per MB for hard disks and for magnetic tape. I double-checked to make sure that I had not miscalculated some of the prices and then looked in my files to find a similar graph compiled in the summer of 1999 – which is shown below as Fig. 3.

Figure 3: Price of storage media as a function of storage capacity (July 1999)

Comparing Fig. 2 and Fig. 3 indicates that most of the points plotted on these graphs have slipped down and to the right between 1999 and 2003. Downwards movement indicates a reduction in price, whilst movement to the right indicates increasing capacity. In 1999, a hard disk with 30 GB capacity was considered to be "very big" whereas, by the end of 2002, the horizontal axis of the graph had to be extended above 100 GB.

In 1999, magnetic tape was much cheaper than hard disk storage: in fact, the price per MB of magnetic tape was about 10% of that of hard disks. As hard disks and tape are both based on the principle of magnetic storage, I recall postulating that the technical advances permitting denser storage on magnetic media (e.g. more MB per unit area) would be equally applicable to both storage media. Consequently, I stated in various lectures that "magnetic tape would remain cheaper than hard disks for, at least, 5 - 10 years". Nobody challenged my statements at the time or since – which is why I am publicly admitting my error.

Why did nobody challenge me? I suppose it was because the statement was self-obviously true! Tape has always been relatively cheap - even if the tape recorders are expensive. Developing new types of magnetic tape requires deep knowledge of physics and chemistry. Manufacturing tape requires sophisticated and expensive facilities. Nevertheless, as tape is made in huge quantities, it can be made quite cheaply. In comparison, a hard disk is much more complicated because it combines high-technology magnetic storage media with precision-built moving parts. The idea that hard disk storage could be cheaper than magnetic tape simply does not make sense .......

Figure 4: Price per MB for hard disks, RAM and magnetic tape

Fig. 4 shows historical trends measured in price per MB. It is similar to Fig. 1 except for the addition of the blue curve showing the price per MB of magnetic tape. Note that there are long periods (e.g. 2 years) of stable or constant prices for magnetic tape. Although the blue curve has a downward trend, its slope is much less than that of the curves for RAM and hard disks. It is very difficult to model the blue curve by a straight line: this suggests that, unlike hard disks, the price trend for tape is not following an exponential law.

Note that the red and blue curves in Fig. 4 cross each other during 2002 – confirming the similarity in price of tape and hard disks. However, looking at Fig. 4, it is worth considering whether hard disks will become much cheaper than tape. As mentioned above, it is important to recognise that the historical price trends may not be valid in the future. Although the price of hard disks can readily be modelled by an exponential law (e.g. a factor of 2 every 12 months), the falling prices have been caused by intense competition between a limited number of manufacturers of hard disks. None of them wants the price of their latest high technology products to drop rapidly, but the purchasers of their products (e.g. computer companies) are very price conscious and, hence, would change to another supplier who can offer a similar product at a lower price. By investing in the latest technologies, manufacturers of hard disks might be able to offer a lower cost product and, thus, gain a temporary advantage over their competitors. It is arguable whether this complex ecology can be sustained in the long term - but, as mentioned earlier, similar fears expressed in the mid-1990s were unfounded.

It is interesting to speculate why tape prices have fallen so slowly during a period when magnetic storage in hard disks has improved dramatically giving consumers much better performance at a much lower price. Perhaps one reason is that new developments in magnetic tape often require new hardware (e.g. video-tape recorders). Manufacturers would benefit from sales of new tape machines, but users (e.g. broadcasters) are reluctant, or simply cannot afford, to change their hardware too frequently. On the other hand, users of hard disks are not interested in the internal workings of the hard disk: if the interfaces (e.g. SCSI, IDE) have been standardised, they are not constrained by compatibility with previous hardware.

Another possibility is that the tape suppliers might have benefited from advances in manufacturing techniques that result in lower costs. In other words, they may be making substantial profits from magnetic tape and see no reason to pass these savings on to the customers. Who can blame them? If that is the situation, the growing competition from hard disks could cause significant upheaval in the magnetic tape industry.

Until now, users of magnetic tape (e.g. broadcasters and company IT departments) have done so simply because tape was much cheaper than any other solution. However, tape has many disadvantages. As all broadcasters know, storing tapes in archives costs real money: apart from expensive storage space with controlled temperature and humidity, complex archiving systems are required to catalogue tapes and ensure that they can be retrieved when necessary (even though most tapes are never taken off the shelf). On the other hand, on-line storage based on hard disks avoids the need for physical handling of tapes by allowing remote users to browse through archives and to select the required items. As users of tape (e.g. in an editing suite) seem to spend much of their time copying material from one tape to another or spooling through tapes to find the required segment, the enhanced facilities offered by hard disks, such as faster than real-time transfers and random-access, are very attractive. Of course, such comparisons must be made with caution. Tape has a long shelf life (if treated carefully) whereas hard disks can develop errors or even fail completely. To minimise the risks of such failures, hard disk systems require complex management systems with built-in redundancy, thus reducing the apparent benefits in terms of cost per MB.

If hard disks really do become cheaper than tape, how many years before we witness "the death of tape"? Although we have all heard people say that "the death of film" is imminent, my personal guess is that tape will be dead within 10 years.

Philip Laven
Director
EBU Technical Department

Editorials Archive