Comment by ksec | Hacker News Reader

Comment by ksec | Hacker News Reader

It is suggesting there are research that shows it is possible to have 20x times the density of current types, which tops at 15TB. But these research are like battery breakthrough, good in theory but not thought through in practice.

While we have a real roadmap, and a working solution of MAMR, much more so than the proposed HAMR from Seagate, that will scale us to 40TB per HDD by 2025, and may be more so given we don't really know the limits of MAMR yet, and we manage to put more platter inside a helium drives. Which I don't think is possible with HAMR, laser heating plate with helium filled drive has got to fun inside labs.

NAND Flash is also dropping in price, the current NAND spot price are already backed to 2016 price or may be even slightly lower. We have Fabs from Toshiba, Samsung and SK Hynix all coming online in 2019, all while 3D NAND are achieving better yields. The "Ruler" that Intel proposed has now standardises and becomes EDSFF, with up to 1PB storage in 1U. All current roadmap suggest we should hit 8PB or 16PB in 1U by 2025. In terms of 48 Drive in 4U server, HDD offers only 480TB per 1U by 2025. That is 16 to 32 times the density difference comparing to NAND.

In a system where information is stored in different drives like black blaze[1] , your reliability reaches a point ( I think, forgive me if I am wrong ) no different to those offered by Magnetic Type. What are the advantage left for using magnetic type? When cost, performance, space, reliability does no longer flavours them?

[1] https://www.backblaze.com/blog/cloud-storage-durability/

Tapes can last for 50 years. They are the MOST stable and reliable form of storage we have.

We have also not even reached any kind of limit on density. Sony has tapes that hold hundreds of terabytes.

So, they last forever, are extremely reliable, and extremely data dense.

HDDs exploit the same concept (magnetic storage) but trade speed for reliability. NAND is more reliable but still has a shelf life, and it also relies on support chips to constitute the overall media. It's also expensive as hell.

Tape is tape. You could smash a cassette and respool the tape and read it. Its also cheap as hell per gb.

Tape will be around for decades if not centuries to come. NAND is not a challenger whatsoever.

A key piece of the puzzle here is that not only are tapes used for backup, but to free up space from data center HDDs in order to increase their shelf life.

Stop thinking of tape as archaic. Magnetic storage is the foundation of modern technology and computing, and in past decades we barely scratched the surface of its potential.

We aren't talking about audio cassettes with analog audio recorded to them. We are talking about a storage medium that is more data dense than HDD or NAND could ever be and we have yet to reach any sort of limit.

With tape you can achieve more and more space by manipulating the width and length of the tape, the tape heads, the angle at which tape is read, and the way that the data is read and stored. It's also just a plastic cassette with polyester tape coated with rust particles. It's inherently cheaper, requires no circuitry, power, or motors within the media itself, and will last 50 years.

Tape is the purest implementation of magnetic storage concepts. As such it is going to improve faster than HDDs because there are no other design factors like platters, motors, and heads. If an HDD dies, you might be able to recover the data through expensive and time consuming procedures. If your tape drive stops working, you buy another drive and still have an intact tape with all of your data.

Additionally, while SSDs are wonderful, they are a pretty backwards design when it comes to efficiency. You are essentially trying to model magnetic storage by using integrated circuits. You are storing values with digital logic in a physical state on a chip.

Its fast and more reliable than a spinning metal disk, but it's expensive and not data dense at all.

The only threat to tape is public perception of it being "old" when it's really the most robust and best engineered data format we have.

> Tape is the purest implementation of magnetic storage concepts. As such it is going to improve faster than HDDs

I agree with the former, but I think your conclusion misses a critical piece of technology development: volume (which drives funding).

Of course, HDD volume is going down, while SSD volume is going up. That may well give tape enough of a comparitive advantage, if it hasn't already.

> The only threat to tape is public perception of it being "old"

For a definition of "public" narrowed a bit to include only/mostly computer professionals, I'd add "slow" as a mis-perception.

Other perceptions I, personally, have is that it's expensive and inflexible, mostly in the context of startup companies. I don't believe that's any kind of threat to tape, however, since it's likely these same companies will gladly over-pay for a cloud version of tape.

Reply on news.ycombinator.com