WHOMP! There it is: IBM demos 154TB tape IBM and Fujifilm have demoed a 154TB LTO-size tape cartridge which could come to market in 10 years' time. While it falls short of the Sony 185TB development, it should, with IBM's tape partnerships, see Sony exit the field. The Sony development involved a 148Gbit/in2 tape density and its own tape design to achieve a 185TB …
"Vulture View: With its LTO connections, such an IBM technology could well banish Sony’s impressive 185TB effort to the tape wilderness. Reckoning on LTO generations doubling capacity we could see a hypothetical LTO-22 pass the 154TB mark."
I think Vulture Centre needs a little bit of work on his mathematics. If capacity doubles every generation, it will take 6 generations to go from LTO 6's native 2.5TB to 160TB, which would make it LTO12 (or LTO-11 if it's compressed capacity at 2:1).
Of course, one of the big problems with a 160TB is just how long is required to read/write the whole thing. Assuming that the bit density is the same in both directions, if you double the linear density, you quadruple the capacity. Even if you double the number of tracks (and therefore read/write heads), you only double the read/write speed so it will take twice as long to read/write an entire tape as compared to 4 drives of the previous generation. Scale up LTO6 to LTO12 and that would, theoretically, mean an uncompressed read/write speed of about 1.28GB per second, or 5.56GB per second on 2:1 compressible data. That means it would take about 34 hours to read/write the whole 160TB. A saving in media space and drives of course, but you create a bottleneck, and all that assumes you can keep doubling the number of heads per generation (unlikely I would say). Basically a bottleneck, not unrelated to that of (linear) write speed on ever larger disks. Namely capacity goes to the square of bit density whilst linear access speed is only linearly related.
(Random disk access is another issue which is even worse).
Yes you're right of course but tape is still being used and will continue to be used for a long time because of compliance issues, of which there are many. Basel 3 & Sarbane Oxley for financial markets for example.
Tapes are easily portable and relatively long lasting especially in the hands of professional data vault companies.
I think that's what the auditors care about more than the restore time.
In any case, most companies these days days keep several "online" copies of the data just in case the data centre goes down. Again a compliance issue.
I'm not questioning the future of tapes for very large archives and very large backups. For those purposes it's still unbeatable for cost, power consumption and robustness in transport. However, what I'm pointing out is that there's something inherently unscalable with both tapes and disks and performance becomes a big issue.
nb. I made a bit of an error with the requirement for more heads. In fact the number of heads is increased linearly with bit density, then total read/write time remains fixed. My analysis only works if the number of heads is kept fixed (tracks have to go up, but that's a different matter). LTO has doubled the number of heads in the past, which explains why the total read/write time hasn't increased as much as expected, but this can't go on for ever. So another doubling of head numbers before this theoretical LTO-12 arrives might allow for total read/write time of about 16 hours.
I missed the Sony article here, but I do remember some articles sprouting up about "Sony bringing back the cassette tape" and me going "lolwut? Sounds like a new LTO cartridge. Those have never died!" And it looks I was right.
I do wonder why tape mfgs love to tout "compressed" data capacities? I remember being bit by this when I was doing my backups on DDS4, only to find out that they didn't fit 40Gbs but 20Gbs unless you compress. Bzip2 was painfully slow on the processors I had on hand back then so I ended up using more DAT DDS4 tapes to speed up the process.
"""A follow-on thought: If tape density can more than double every 30 months then it could well outpace disk density improvements and cement its role as the archive medium."""
* Tape storage halves in price per GB every 30 months
* HDD storage halves in price per GB every 18 months.
* Flash SSD storage halves in price per GB every 12 months.
So at what point will HDD take over from tape?
At what point will SSD take over from HDD?
At what point will SSD take over from tape? (Possibly never to this one. Currently, high capacity SSD needs to be powered up every few months).
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
