NAS door gives drag-n-drop access to linear tape vault

With StrongBox Crossroads has built a NAS head for tape libraries that combines disk access speed with tape's low-cost and longevity, and can cut file storage costs by 90 per cent. That's Crossroads' boast, although the NAS head is our term, and Rob Sims, Crossroads' president and CEO clarified it by saying: "It's a NAS head …

[CheesyTheClown]

Cut cost by what?

I run a 14 terabyte raid at my house... The total cost of the unit was about $1500 and that was long ago. I can mirror the RAID with rsync (or equivalent) to another location quite easily providing 100% redundancy. I can replace the drives in the mirror with 4tb drives instead of 2tb drives and even have revisioning.

The cost of running the system is painfully low... the machine has a total average power consumption of 27 watts and successfully floods the three internal gigabit ethernet adapters. The precise same computer build has been used for full 4K video production on 5 Avid, Final Cut Pro and Adobe workstations where non-linear editing was performed without a hiccup.

Could I implement the same tech on a large scale (petabytes a dozens of gigabit ethernet links)? sure... no problem. I could easily scale this tech to petabytes or tens of petabytes with full redundancy. Mirroring would even be easy using point to point 10Gbe adapters. The trick is simply using file systems which support distribution.

Tape technology is nice, but it's way too small, way too slow and way too unreliable. Yes, unreliable. Tape is a shitty solution these days. You're making use of an insanely long strip of magnetic media which is being spun in layers in a harsh environment around a spool and then transported to a "safe place". Error detection and correction improves it a bit, but the bit density of tape is just too high to reliably store media as such... therefore you need multiple tapes of the same data.

It's just too slow and too unreliable for modern use. Multi-location raids are far more reliable and, if implemented correctly, power friendly. It's also substantially faster.

My dream at the moment is a single board computer with a ULV Core i5 (such as those in tablets), 16GByte soldered on ULV DDR3, 4 accelerated Gigabit ethernet ports (including jumbo packet support and IPv6 TCP/UDP checksum calculation) and a soldered on HighPoint-Tech 8-port SAS 6Gbit controller with support for SAS expanders. Oh... and full support for low power on all accessories when idle. A soldered on 128Gb boot drive would also be awesome.

A board like this would probably have an average power consumption of 8watts. It would peak around 40. Using 5400RPM drives with this board would have an average power consumption of well under 2 watts per drive each.

There's no point wasting time with a multicore ARM processor as nearly all the power is likely consumed by the peripherals and frankly, unless highpoint makes their own ARM cpu, it's almost certainly going to be some shit controller from some lame ass ARM company who focused more on the ARM than on the RAID.

[Anonymous Coward]

Hmm...

This system isn't targeted at you, it's targeted at really huge enterprise environments who pay knocking on towards a million quid for an array pair.

Just have a look on the Internet at the IBM libraries that they listed as compatible and you'll get an idea of just how big this sort of system is.

Also for something like your use - think about how you'd store a feature film after the final cut, if you're doing non-linear editing? It's not staying on the disks, but you need it somewhere, a tape hosted filesystem is perfect.

[Andrew Downes]

Or you could investigate a tried and trusted solution...

IBM Tivoli Storage Manager for Space Management does the hierarchical storage bit (which is what integrated disk and tape means).

It also integrates backup which is somewhat important - copies of the tapes in the library, copies to take off site, even a fully automatic duplicate archive at a second site if you need it. Try doing that with file-by-file backup.

It has a worldwide client base including Soho post-production houses going back many years. Enterprise scale and reliability. Teamed with V7000 Unified can offer multi-tiers of disk (high performance 15k RAID10 to near line 7.2k SAS).

But fundamentally the software is disk independent and supports a massive range of tape libraries, does not depend on LTFS support.

Full disclosure: I work for IBM and get a little frustrated when I hear clients spending more than they need to to on "new" function that has been in production for years if they just asked (in this case) the global leader in tape technology. Doesn't have to cost a lot, try us.

[Anonymous Coward]

Hmm...

Isn't LTFS an IBM technology and also free? (I can't remember, but I seem to remember having heard that it is) Also, how do I move all my data out of TSM when I want to move it to another backup/archive package, how much extra hardware do I need to attach to my TSM environment and how many more TSM client licences do I need to run an archive system like this?

[jubtastic1]

RIS?

If they're selling the storage based on a 29K average file size how on earth is a backend tape library supposed to get to the last 14K before a disk has served the first 15K?

[Anonymous Coward]

Well, they mentioned SAP. I guess that makes another $31,000 for storing 14TB seem almost reasonable in comparison to the millions their poor customers will have spunked on a failed SAP implementation by this stage...

[Anonymous Coward]

Let somebody do it for you?

I can't say that buying the hardware yourself, having the running costs yourself and then having to worry about the data yourself seems like an ideal solution to me.

Staff and their time are not free.

There are plenty of other options out there, including at least one company that offers a service which includes local disk cache, with the data backed off to a tape based cloud services and given data integrity guarantees too.

If you're not so bothered about data integrity, then there are the usual suspects to look at too.