Do we need a new tape standard?


Userlevel 7
Badge +13

The question may sound provoking. I asked myself during reading an article about Quantums concern about the LTO roadmap:

https://blocksandfiles.com/2022/06/17/quantum-thinks-lto-tape-falls-short/

For short: Capacity of tape media fall behind capacity of single disks. Also therefore tape does not fit for hyperscale that good. They need huge volumes of cheap storage. Best solution would be to use tape as very slow disk drives.  And because hyperscaler use tapes within their own environment, they do not care about standards. And because they are the biggest customers for tapes, it could happen that we see a new tape technology arise. At least this is my option.

What do the community think about this?


8 comments

Userlevel 7
Badge +17

The smaller particles and the different chemical combination should not influence the duration. The important point is that the tape does not touch the read/write head or another “flat bed”.

The speed the tape is moved increases with every generation, so we have a tension increase because of this. At the same time the tapes get thinner.

So the tighter winding would be an additional factor for the mechanical strain of the tape….

Userlevel 7
Badge +20

I agree, a wide spread standard is important. The 3592 are excellent, but proprietary and only supported in the big IBM Libraries. All other existing standards except LTO and 3592 are dead in my opinion.

So, if there will evolve some nice new standard I will gladly accept and test it. 😎

 

But please… no higher tension or physical touch to the tape. We had this with several other standards and this killed the tapes in very short time.

 

There is a new chemical combination tested which could help with higher write density due to better magnetic properties and smaller particle size: Strontium Ferrite (SrFe). At the moment Barium Ferrite (BaFe) is used. Nowadays the BaFe particles have a size of 2000 nm^3, getting smaller to 1200 nm^3 with the next generation. With SrFe particle sizes of 900 nm^3 (and probably less) are possible.
We will see it this is a solution for the higher capacity demands.

I hereby crown thee: Sir Tape Master

 

Is there anything you don’t know about tape? 😆

 

Taking this comment to a particle discussion brings back around the question of tape tension. Are the new tapes going to be as physically durable and operate within the same operating conditions, as we may gain in density but lose in some other metrics that dampen the benefits. We saw this with Apple and the Mac Pro prior design to current, aka the “trash can”. It was a great looking design, but with the trends of computing still relying on power draw increases to gain more performance, it was thermally constrained quickly.

 

Be a good space to watch!

Userlevel 7
Badge +17

I agree, a wide spread standard is important. The 3592 are excellent, but proprietary and only supported in the big IBM Libraries. All other existing standards except LTO and 3592 are dead in my opinion.

So, if there will evolve some nice new standard I will gladly accept and test it. 😎

 

But please… no higher tension or physical touch to the tape. We had this with several other standards and this killed the tapes in very short time.

 

There is a new chemical combination tested which could help with higher write density due to better magnetic properties and smaller particle size: Strontium Ferrite (SrFe). At the moment Barium Ferrite (BaFe) is used. Nowadays the BaFe particles have a size of 2000 nm^3, getting smaller to 1200 nm^3 with the next generation. With SrFe particle sizes of 900 nm^3 (and probably less) are possible.
We will see it this is a solution for the higher capacity demands.

Userlevel 7
Badge +20

Disclaimer: I don’t have a preferred technology in the runnings here. I was using USB removable drives in my earlier data protection days, and I see the value of both tape & cloud storage.

 

I agree with @JMeixner that tapes are designed around longer term retention, so when I see comments in the article such as increasing tape tension, I can only assume that would either have constraints around longevity of the media, or alternatively, more strict storage requirements to preserve the tape.

Tape has something that is missed on all competing solutions presently, mechanics. The ability to eject a tape physically, in an automated fashion, providing no ability to write to that tape without manual intervention, is a brilliant defense against ransomware. Software defined solutions can never match this, be that kill switches on switches etc, it is all relying on a software approach to disrupt the connectivity, and it’s always talking to a far end, instead of an integrated component.

 

The ability to write across tapes has always helped protect its scalability, and I do certainly hope we see increased tape storage density, since archive storage tiers in the cloud are just writing out to tape anyway and charging huge amounts to recover from this, I do like the idea of local tape still!

 

I do see a potential benefit to a new tape standard, if it drives innovation and competition, it’s certainly been a pain seeing the growth slow, but we need widespread standard adoption or we end up with fragmented support.

Userlevel 7
Badge +17

Yes, LTFS is available. Can be used with Spectrum Archive for example.

Depending on the application managing the tape this gets you some improvements to access a specific file on a tape. But if the managing application records the positions of the files on the tapes it not that much…

The advantage of this that there are caches and not every file has to be read from tape as a single operation.

 

Super is to use this as a abstraction layer to access the tapes from several application without noticing them that thy are accessing tapes.

Userlevel 7
Badge +13

Can not fully agree with you, @JMeixner ! Have never used it, but still there is something like a filesystem for LTO: https://en.wikipedia.org/wiki/Linear_Tape_File_System

Userlevel 7
Badge +17

Mhh, regarding the article the hyperscaler want to use the tapes as slow hard disks. For this the tapes have to be present into a drive.

I think, this is not the intended use of tapes. tapes should be offline in the library slot or even better offsite of the library to be air-gapped. Then the usage of tape provide additional security and saves power.

 

As for the LTO roadmap: LTO seems to be still recovering from the LTO-8 disaster. We will see if LTO-10 doubles the capacity or if these times are over…
Jaguar (IBM 3592) tapes with 20TB capacity are available since 2018, the next step with tape capacity of 30TB should be in the very next future…

And we have to see if the HDD manufacturers will really release HDDs with 40 or 50TB capacity. These are concepts up to now - as far as I know. Tapes with more than 100TB capacity are in real world test now, they are not theory only...

Userlevel 7
Badge +20

I think tape will stick around but am beginning to wonder about VTL picking up maybe for other scenarios where tape does not fit. Who knows but we are adopting more and more tape for specific services we have for our clients as deep storage and air-gapped.

Comment