Walking the floor of the SNW-Europe expo mostly meant seeing companies and products already familiar in the US. But there was one surprising exception: Fast LTA.
The LTA stands for Long Term Archiving. FAST LTA AG, based in Munich, is headed by founder Matthias Zahn, and is his 3rd startup.
Founded in 2007, the company’s main product is called Silent Cubes. Each stylish silver cube contains 12 drives, either 2.5″ for the personal cube, or 3.5″ for the corporate cube – which requires a management head-end – and can survive 4 drive failures without data loss.
The 12 drives come from 3 different vendors to reduce the risk of epidemic failure. The controller – either head-end or internal – implements the NFS/SMB interface and what the company calls “hardware WORM” so archived data can’t easily be overwritten.
Here’s a diagram of the product:
The Cubes are half-rack width, so 2 can be mounted side-by-side in a 19″ rack. And they only consume 2 watts at rest, waking up once a month to scrub the data and check disk drive health.
The StorageMojo take
StorageMojo has asked disk drive vendors for archive-class disk drives for years and who knows, if they get desperate enough maybe they’ll decide there could be a market. In the meantime we now have Silent Cubes.
While the concept is good, there are several things that Fast LTA could do to improve the product:
- Move to a more advanced form of erasure coding – they use Reed-Solomon – to get more capacity or even greater failure resistance.
- Get serious about the US market. One question: under the Federal Rules of Civil Procedure, is data whose index is destroyed but the data is not – which is how Silent Cubes “deletes” data – still subject to production during discovery?
- Put some color pictures of that nice design on the corporate web site. Why spend money on design if you’re not going to show it off?
Fast LTA would be a logical launch company for shingled drives if their volumes are sufficient. But there is clearly demand for archive quality disk storage.
Are drive vendors listening?
Courteous comments welcome, of course. I was at the show thanks to Amplidata, which makes a large, scale-out active archive system that I like.
After reading the excellent explanation on http://www.pdl.cmu.edu/ftp/News/newsletter11.pdf, it occurs to me that shingled disks as proposed require different integration techniques for different applications and operating environments. That means a fragmented market and makes it look like a non-starter. Those problems would be avoided if the single track head is replaced with a multi-track head, where the subsequent write-heads are staged so the head writes a shingled band in one go. This would give the increase in areal density without the severe system-level penalties and allow these drives to function as drop-in replacements for all environments. The key question is of course if the increased density would overcome the cost increase for multi-track heads. When factoring in the full cost of a storage system per drive, there is ample budget for these heads.