Rather than using an ‘archive quality’ disk (as defined above), one alternative is taking industry-standard disk & leveraging MAID techniques to achieve a similar end-point.

Nexsan is one vendor that has enhanced MAID (AutoMAID) that’s interesting in practice (sales disclaimer – don’t work for Nexsan, but as an IBM specialist, we also do use Nexsan on occasion). We can use bulk MAID disk within our infrastructural approach to building enterprise-wide archives.

In projects requiring a persistent, ‘active’ archive, we’ve found that a storage grid topology (using IBM’s GPFS & TSM), creates a scaled-out tiered storage approach, offering the option of abstracting tape (via TSM’s HSM) as an underlying storage tier (essentially using the guts of SoNAS in a custom, purpose-built approach).

Rather than the clustered NAS approach (file system closed within the walls of the array), exposing the GPFS filesystem allows us to loosely couple applications (transcode, encryption, etc.) to the compute (hpc) side of the grid, while offering scale-out performance/capacity, and thanks to TSM, built in data protection for backup & DR. …ie, no NDMP-based backups (which don’t work well in PB-sized environments).

Apologies for the sales pitch, wanted rather to communicate an infrastructural approach to solving real-world problems.

Best -

John Aiken

Thanks for the update.

Regards,
Jim

“Archive quality disk” – I didn’t invent the idea, but maybe the term. A disk you can put on a shelf for 10-20 years and still read. Requires different lubricants, probably 5400 RPM or less, extra-robust ECC, maybe HAMR, native USB interface or a little dock. ProStor + shelf life.

Data longevity trumps performance and IOPS. I’m told Seagate built a disk for car navigation systems guaranteed for 10 years.

Sell ‘em to consumers and SOHO/SMB by the millions – anyone who can’t see laying a few grand for LTO. I’m sure there’s someone in one of the company labs itching to try this. People are buying those USB docks for a reason. With USB 3 they will have good bandwidth and a long interface life.

Robin

Archival storage is a huge problem, and one that is only getting exponentially bigger as time goes by and data retention requirements go up. But like you can’t fight the next war with the last war’s tactics, I doubt we’ll solve the 21st century’s storage problem with 20th century ideas. $0.01 per Terabyte, 25 year retention, 100 GB/sec transfer rate, zero static power loss, and time to recover on the order of minutes rather than days. I don’t think such a system exists yet but I know that if you think you’re designing an archival storage solution you should measure your fitness against that particular razor.

–Chuck

It was pretty clear to me from the beginning that given even an optimistic reading of InPhase’s roadmaps–and an optimistic reading of a roadmaps is, of course, always highly charitable–that while the technology was interesting for technology’s sake, there was no really good use case for it.

“Archive quality” disks would be an interesting subject. Not sure what it means, yes, someone please say more. I may feel impelled to go get NDA roadmaps briefings from the relevant manufacturer.

C//

Your parting comment does interest me (due to limitations with current magnetic media)… “archive-quality disks”. Any information I could read on the topic? Thanks – Jim

Still as you pointed out, InPhase’s product became really obsolete due to the advance of the current hard drive. I truly hope that someone will learn from this “experiment” (because unfortunately this is what it ended up being despite the good intentions) and reuse/move the technology forward.

Thank you for sharing.