Slashdot Mirror
Holographic Storage Crams in 0.5TB Per Square Inch
An anonymous reader writes "VNUNet is reporting that a company called InPhase Technologies claims they have successfully recorded 515GB of data per square inch to capture the record for highest data density. From the article: 'InPhase promised to begin shipping the first holographic drive and media later this year. The first generation drive has a capacity of 300GB on a single disk with a 20Mbps transfer rate. The first product will be followed by a family ranging from 800GB to 1.6TB capacity.'"
515 Gb is only 64 GB. So about 4.6 square inches of data surface on a 300 GB disc.
Why would anyone engrave "Elbereth"?
The summary incorrectly lists 515GB, but the article says 515Gb...
With a 300GB capacity and 20Mbps tansfer rate, it would take 34 hours to read or write a single disk. Assuming they made a mistake in the transfer rate and its actually 20MBps (possible though unlikely considering HD-DVD drives are shipping with 35mbps, or ~4MBps rates), it would still take ~4 hours to transfer a disc. I can burn a 700MB CD in 5 minutes, and a 4.7GB DVD in 25 minutes.
We do not live in the 21st century. We live in the 20 second century.
https://www.eff.org/https-everywhere
According to this video, the media can be stored for 50 years, and it also looks more like a replacement for tape backups than hard drives at this point.
Coupled with this article, which says that it's "10 times faster than a normal DVD burner (whatever that means)," and holds about 300GB (278 GBytes formatted) it's clear that they're aiming for removable media.
Apparently each 300GB disk is about the size of a DVD (but thicker due to it having it's own little shell, like a floppy/zip/mini disk). Just like all removable media, it needs its own drive. Unlike most, it needs a HUGE drive, about the size of one of the old tape drives (2xCD-ROM drives but longer).
It's an interesting backup solution, maybe if you got a bunch of these and made a RAID-0 across them, you'd have something really cool.
Latewire
I think it's just the new higher 'resolution' technology that's presently limited to 20MB/s.
The older stuff that got something like 200G/in^2 had a much faster transfer rate, but lower density.
It's all due to the electronic end, really. The laser picks up the data at 1G/s, but the electronics take much longer to send it out to whatever bus reads it.
110100 1101000 1101000 1100110 0 1101111 1101000 1100011 1
It's not a HDD replacement. It's a CD/DVD replacement. So imagine something that looks like a floppy disc holding 300G of data.
Also remember that this is the first product to use this technology. In a few years we will look back on this and think about how amazingly slow it is, and how slow it is.
Yes. Someone has problems with bits and bytes.
20Mb should be 20MB
515MB should be 515Mb
No, not three dimentions.
The idea behind a hologram is that you can encode information from a higher dimentional space to a lower one, without much loss of generality. This device records three dimentional phase information (from the laser) onto an effective two dimentional subspace (the diffraction grating).
The initial InPhase Tapestry(TM) holographic recording device will record 300 gigabytes (GB) of data onto a 130 mm disc with a transfer rate of 20 megabytes per second (MB/s).
Your question is really backwards ; the idea of writing the same data to two or three places is just a very crude form of the "normal error correction" you refer to. "normal error correction" on CDs is implemented by cross-interleaved Reed-Solomon data encoding. It is far far more efficient than simply duplicating the data several times.
If you can describe the error model of the medium, that is what types of errors are likely to occur (random dropouts, scratches which cause burst errors), you can then lay out the data on the disk to convert the statistically most likely errors into error distributions that your error correcting code is most able to deal with. For example, the Reed Solomon code deals with random errors the best, so you use two dimensional interleaving to convert burst errors (caused by scratches) into random errors sprinkled over many code words.
The Reed Solomon code can correct an unbelievable number of errors, it is almost perfect. If you have 2N parity bytes, it can correct N random errors or up to 2N "erasures" (errors where you know the location of the error, due to the underlying demodulator telling you something is fucked).
So if you naively just duplicate 100 Gigabytes data three times, you've got 300 GB but then you're screwed if two bytes of that data are corrupted in the different copies. If you use Reed Solomon, you would only need two extra bytes to give you the same protection.