Tape versus Optical Storage for Archive

by Marc M. Batschkus

In the video industry, there is a long history of using optical media for recording and archiving clips. This is due to the fact that years ago, CD-R and later DVD-R were the only media with sufficient capacity and an attractive price point to store video files for long-term. Additionally, camera vendors integrated optical storage in their cameras. This has not been the case in other industries where optical storage is rarely found.
For many, optical storage seems like a realistic option today to archive their files. Since archiving files implies a long-term perspective, a closer comparison of tape and optical technologies is necessary to take a responsible decision.

Capacity 
Since optical media are very limited in sheer surface area, several media are combined in cartridges to increase total capacity. Tape has a considerable surface area since several hundred meters (960m with LTO-7) are coiled on the built-in reel. This alone shows that the race for maximum capacity can probably never be won by optical media in its current form. Additionally, enterprise tape drives most of the research efforts and products have already reached capacities of 8.5TB in T10000D tape and 10TB in TS1150 tape). These technology advances will be carrying over to LTO tape.

Error rates
All storage technologies have theoretical error rates that can hardly be grasped by a normal brain. Whether it is 1 bit in 1 x 10E17 bits (error rate of LTO tape) or 1 sector in 1 x 10E14 bits (Desktop SATA drives), what does that actually mean? (specialists might point out here that the error rates are in favour of tape even with identical numbers since with tape one bit in …bits is lost where with disk one sector in … bits is lost) For most users and uses, bit error rates in both cases qualify as “reliable”. (this changes when you wear the hat of someone in charge of many Petabytes of data)
Real world reliability comprises many factors and, when those are taken into account, a very different picture can be seen. One example is replacing the often cited MTBF (mean time between failure) with a MTTDL (mean time to data loss) which yields interesting perspectives. Reliability of interfaces, complexity of setup and components, real-world field use and experience, and software bugs all contribute to the complete picture.

Performance
Due to the linear nature of tape, access time to a single file can vary from seconds to minutes. Optical storage has advantages here since it offers random access. When accessing smaller files, optical storage has a performance advantage. Both might need to position the respective media in the drive first – no advantage here for both.
Writing performance of LTO-7 tape is tremendous with 300MB/s uncompressed. It can even be multiplied with drive parallelization that some professional archive solutions provide (like Archiware P5 Archive). Optical storage only offers a fraction of this read/write speed and there seems to have been only little improvement over the years.

Long-term experience
There are claims that optical storage can last 50 years or longer. This can only be simulated since current optical media have not been around for long enough. If this is true market volume and vendor choice are crucial to actually make use of this possible advantage in decades from now.
Magnetic tape has been in use from the 1950s and has been shown to last for decades. Since there are multiple vendors and a big number of drives and libraries of all generations in use, chances are much higher to actually find a device to read a tape in many years from now.
The whole discussion of long-term storage perspective needs to focus on data migration that needs to take place every 5-10 years to be able to use data in a current workflow. No one puts media away for ever and no storage is independent of developments in other areas like operating systems, interfaces etc. For LTO and enterprise tape there are regular upgrade cycles. New LTO tape generations always offer much more capacity and speed that make a migration attractive and economic.
There is one more long-term consideration concerning re-use of media: after media migration, like from LTO-5 to LTO-7, you can re-use all tapes either for a different archive or for Backup since there are many Backup solutions supporting tape. Optical media cannot be re-used for backup at all and hardly used in other applications since the choice is so limited.

Vendor choice
Especially for a long-term archive, it is crucial to be able to rely on several vendors supplying the needed devices and media. Any one vendor might change focus or stop building products. With tape, there is a number of vendors that build devices and media. Since sudden changes in IT used in data centers and big corporations are not to be expected, future supply of tape technology is secured for at least a decade (and that is a lot nowadays).

Overall market adoption and volume
Market size and volume drive R&D and also availability of products and their distribution. Here, tape has a big advantage as there are a number of vendors. Drives as well as media are available all over the world. Options for buying bulk reduce cost per gigabyte even further. New generations with more capacity and faster performance are built to satisfy the broad existing market where about 80% of all data centres use tape storage today.

Summary:
Optical storage today has lost most of its attractiveness of years ago. Price per gigabyte, capacity, vendor choice and market volume all turned around in favour of tape. With the limited exception of seek times for small files, performance of optical storage can not compare to current LTO and enterprise tape. Since most archived files will hardly or never be accessed anyway, access time is of minor impact. Considering the fact that enterprise IT, finance, insurance and other industries all use tape technologies, results in market volumes, demand for growing performance and capacity etc. that will never be there for optical storage. Real-world long-term reliability for optical media can only be claimed but not demonstrated since current optical media have not been around for long enough. It seems that only a very limited number of scenarios are left where optical storage for archiving might be a good fit. This might be the case if there is already an established infrastructure and support for a specific workflow. Even then, the time frame to consider for archiving should be only mid-term to avoid possible data loss and to anticipate market changes that might limit the options of optical storage even further.
Tape has all it takes to fulfill the needs of modern archiving: capacity, performance, scalability, choice of vendors, software support, interfaces and devices, longevity and best cost per gigabyte. Any professional environment considering true long-term archiving should have a closer look at tape.