- casino online slots

LTFS: The Beat Goes On

by Administrator on June 26, 2012

With power flickering on and off as Tropical Storm Debbie whipped my place in Florida with wave after wave of wind and rain this weekend, I found myself on the phone doing scheduled and unscheduled interviews with two genuine experts in tape technology, Matt Starr CTO of Spectra Logic, and Rob Simms CEO of Crossroads Systems.  Thanks, in advance, to both men for sharing some of their personal time to chat with me about the peculiarities, nuances and context of the Linear Tape File System (LTFS).

With a couple of weeks between me and IBM’s Edge Conference, I wanted to hear other voices about the promise and practicality of LTFS, a technology that I am hoping will help us to realize the capacity-per-watt-optimized storage platform that is ideally suited to store seldom accessed user files en masse without costing a lot of money or consuming a lot of energy.

I have since written a two-column arc for that will drop in the next week or two that incorporates what the experts said, plus some great insights offered by the folks from the FujiFilm Recording Media/FujiFilm Medical Systems joint-venture “cloud archiving service” — Permivault — that uses LTFS-enabled storage today.  I hope everyone who shares my interest in this technology will have a read.

To summarize the good news/bad news aspects of the columns:


LTFS is real.   But the code itself provides only part of what you need to build the platform I have in mind.  To get to nirvana, we need to build on top of LTFS using a media asset manager, or an archive management system, or a tighly coupled file system namespace.  Otherwise, LTFS is just an interesting way to attach some storage in a manner that mimics a USB key.

Connecting the roof to the foundation is being done today, but the emphasis seems to be on the use of tape for deep archive rather than as a respository for semi-active data.  I don’t mean to split hairs, but archive is not the same as TapeNAS, which is what I want and what I think will resonate with end users horizontally, across industry verticals.  I know, I know:  the Active Archive Alliance is trying to bridge the gap between archive and active storage to create a new category.  That’s all well and fine, and probably just as accurate a description of the TapeNAS repository as what I am advancing.  But TapeNAS doesn’t use the term “archive,” which tends to put people off who see archive as an enormously difficult undertaking that requires resources and time that they just don’t have.  I would like to see TapeNAS used to describe these mass storage platforms because it uses familiar terminology and connotes a plug and play storage platform without the luggage — positive or negative — about archive.

Considerable success is being achieved with the sale of these kits to media & entertainment shops, audio/video pre/post-production houses, video surveillance (is that a vertical market?), but those guys get tape and deal in files that need to stream.  They think of archive differently than your standard run-of-the-mill commercial data center shop.  Go ahead and use archive (or the sexier “media asset management”) in your meetings with Hollywood.  It means something there.  For the rest of us, talk about building a 35TB mass file storage platform costing no more than $26K with a capacity of 30+TB today, growing to 10x that capacity when Barrium Ferrite cartridges hit the street all powered by less energy that what is required by a five disk RAID array.  That will turn up the volume.


Before you can sell such a kit, however, we need to get over a few hurdles.  IBM seems concerned that someone is going to rip them off — that their “one and only true LTFS” will become Balkanized by the likes of Oracle/Sun/STK if it isn’t formulated as a standard with some sort of certification program.  Unfortunately, they have decided to use “the SNIA” as the court of record for the work.  I find this to be a profoundly questionable choice, given SNIA’s past record in standards dev. (I know, I left off the “the” before SNIA — they hate that!)  Recall SMI-S:  good idea at the beginning, but perverted and ultimately watered down once it fell prey to SNIA-ite politics.  Also, to my knowledge, there are no certification bodies ensuring compliance to any standard.  Just plug fests.  Lots and lots and lots of plug fests.  Going this route might deliver a standard, but it will be quite a ways off in time and we need TapeNAS now.

Device drivers are also an issue.  IBM recommends, of course, its FUSE driver to be used in concert with LTFS.  However, using a proprietary driver set in connection with an open standard seems as silly to me as the way VMware supports RESTful management:  their REST works, but only if you code to a half  dozen proprietary APIs exposed by the hypervisor vendor.  That ain’t RESTful management.  We already see TapeNAS (aka “Active Archive”) advocates like QStar Technologies writing their own device drivers.  I imagine that this trend will continue.  Does it interfere with the valid objective of creating a universal LTFS format with guaranteed support for tapes recorded on anyone’s LTFS kit?  It just might.  So some accord needs to be developed that will allow vendors driver flexibility without corrupting the exchangeability of LTFS formatted tapes from vendor A’s box to vendor B’s.

Finally, there is the issue of vendor enthusiasm to field LTFS mass storage platforms at all.  For vendors already offering disk systems, I wonder how they will regard the “threat” posed by highly capacious and affordable tape solutions for active data.  I had many conversations at Edge with IBM sales droids and channel partners.  Seems like (a) hardly anyone knows about LTFS and (b) their initial worry is that it might cannibalize some of their disk array sales.  Depending on how margins are configured, disk centric salespersons may not like the idea of offering a tape-based mass storage solution, even if it makes sense for the customer.  Let’s not dismiss this potential obstacle to LTFS adoption:  as with government laws, the willingness of the bureaucracies to actually implement and enforce them is the ultimate determinant of success.

For now, with the exception of what I have seen with Crossroads Systems’ StrongBox TapeNAS head, which preintegrates all of the components necessary to put a TapeNAS into play today, everyone else seems to be offering a science fair project approach that DIYers may like, but that certainly lacks appeal to the mainstream IT guy or gal who has no time to cobble together drivers, LTFS software, a head server, and an interface and to perform all of the testing to ensure that it is stable and “enterprise ready.”

Watch this space.


Previous post:

Next post: