StorageMojo

Over at Forbes, Seagate Technology Chief Executive William Watkins talks about the future of disk drive capacities and pricing. The money quote:

My best guess is that you will get a 40 gigabyte drive in 2010 for $80 on flash. I think the same device in a one-inch hard drive will cost you $40 to $50. In a 2.5-inch drive for $50, I suspect you can get 400 gigabytes. I don’t see flash being competitive at 40 gigabytes. But I won’t be able to do a 20 gigabyte for much cheaper than [the 40 gigabyte drive], while they will be able to do 10 gigs for $40. Anything below 20 gigabytes, they will own.

. . . Our sense, especially with application sets that want to do a lot of streaming video, is that the minimum requirement will be 50 gigs with a more likely minimum at 100 gigs.

A Close Reading of the Tea Leaves
What is interesting here, four years in the future?

• Flash at $2/GB, 1″ hard drive at $1/GB
• 2.5″ drive at $0.12/GB
• The applications he’s talking about aren’t laptops, which are close to those capacities today, but media players.

Dream On, Mr. Watkins
Flash prices today are only 2.5% of 2001’s prices. Since flash prices are driven by semiconductor technology, primarily larger wafers and smaller features, that trend will continue. So today’s $20/GB flash drive will be a $0.50/GB drive in early 2011.

Assuming Mr. Watkins is correct about hard drives, late 2010 will look like this:

• A 200GB 2.5″ drive will about $40 (applying his reasoning about 1″ drives)
• A 100GB flash drive will be about $50
• Ergo, flash will own everything under 100GB

Remember, you heard it here first. And please, dissenting opinions welcomed.

Wouldn’t be great to stroll into an internet cafe, stick a flash drive in the USB port, and have your desktop, bookmarks and documents immediately available without leaving any traces on the host machine?

That is what Lexar is promising to ship in July with PowerToGo software on high-end 1 and 2 GB flash drives. The software comes from Ceedo Technologies, an Israeli software company, that says

Ceedo Personal is a powerful working environment for portable storage devices such as USB flash-drives and hard drives. It allows users to carry their programs, documents, e-mails and favorite browser with them, accessing them through an easy-access menu once connected to a computer.

The Ceedo [OS] provides Ceedo users with an independent, private working space, created separately from the host computer’s operating system. Ceedo users can run their carried programs on any [Windows] computer, work on their documents and access their e-mails on the go.

You don’t have to wait for Lexar. Go to Ceedo, download the software, and try it yourself.

Be aware that tiny hardware key loggers can capture passwords and messages on public PCs. Use an on-screen keyboard and the mouse for passwords. If the security of your messages is important, stick with a laptop.

You can also load “portable” applications on a thumb drive. Anyone competent to comment on the security of that?

16GB Flash Drive Price Begins Downward Slide

The price of the 16GB USB 2.0 flash drive dropped to $1175 at mWave, an almost $200 drop from less than a month ago. Two data points do not make a trend, but let’s watch this and see if a 15% drop per quarter trend develops.

The online low prices for flash drives tell an interesting story of where the sweet spot for flash is today.

The 4GB flash drive is the sweet spot today.

While a nightmare for security people, very handy for people who don’t have or want a laptop to lug home from the office. Load your documents file into before going home and you’ll have all your work handy on your home machine. Aren’t we getting lazy?

The emergence of a 16GB USB flash drive for $1363 poses an interesting question: will Samsung’s pre-announced 32GB flash drive cost $2600?

Clearly there is some kind of kink in the price curve when a 4GB flash drive costs $19/GB, 8GB costs $32/GB and 16GB version costs $85/GB. Maybe this pricing is designed to skin early adopters recover design costs for the larger packaging. Or maybe they are using the latest and priciest chips, since chip cost is usually about 90% of the device cost.

I’ll be very surprised if the 32GB drive comes in at much more than $20GB. Of course, if Apple is the first customer and its an add-on, they’ll price for big margins. Expect a $1400 price from them.

With Samsung’s announcement of a 32GB hard drive replacement using flash memory (like that in USB thumb drives) instead of rotating disks, the SSD is set to conquer a new market: the ultra-portable laptop. It is about time.

I owned and daily used for over five years the first SSD-based laptop: the original HP Omnibook 300, which was also the first usable ultra-portable laptop. The specs are laughable today: a 20MHz 386SX, 2MB RAM, 10MB flash drive (which replaced the 40MB HDD for an additional $400) and a 9″ grayscale non-backlit screen. ROM software including a usable PDA, MS Word, MS Excel, Laplink, MS-DOS 5.0 and Windows 3.1. The whole thing weighed less than 3 lbs, and with the flash drive the battery life was over 9 hours, which meant I’d only recharge it every few days. Plus, closing the case put it to sleep and opening it actually woke it up! It went weeks without re-booting Windows - a tribute to HP’s engineering. The sleep function has never worked for me on any other Windows laptop including IBM, Dell and Sony. And the 300 had a very cool pop-out mouse that worked well and made people ooh and ahh.

The benefits of flash-based disk are:

• Power draw is dramatically reduced over a hard disk.
• More reliable than hard drives - remember the camera flash cards recovered from destroyed cameras?
• Lighter weight.
• It looks like Samsung is fully realizing Moore’s Law by doubling - or more - flash memory capacity every 18 months, so 32GB will soon become 64GB, rivaling (at greater cost, of course) current 1.8″ drives

Naturally there will be some unexpected consequences of the 32GB flash drive:

• Flash disk based ultra-portable laptops (and I expect Apple will be first to market with one given their close relationship to Samsung) with long battery life will start taking away market share from Treo-like devices. Why have a dock when you can have all your work with you?
• Notebook disk drive prices and margins will drop. The 2.5″ and 1.8″ drive vendors now have an alternate technology that will give them a run for their money while also forcing them to pick up the pace of capacity increases.
• Backup software will need to change to reflect the reliability of flash. Failed disks will basically be a thing of the past, so backup vendors will need to focus on issues of data corruption and theft. And that means losing the fussy interfaces: you plug into your office dock and all your data is replicated up to the second you unplug, and the user does nothing.

Samsung’s entry into the high-capacity flash disk market will be great for consumers as long as vendors design for battery life and not just light weight.

Slash dot points to this story from the BBC and AP about military secrets being sold along with USB drives in Afghan bazaars. The flash drives are stolen from US military bases by Afghans working in them.

A few thoughts:

• Maybe they should pay the Afghans a little more, and then fire a few for theft, to get folks to be concerned about keeping their jobs.
• Why the hell are sensitive files being kept on flash drives? Aren’t they are on reasonably secure network servers?
• This is exactly the same issue that US businesses face. Your entire customer contact list could walk out the door a dozen times a day and you’d never know until competitors are calling on them. This is the real data security problem, not social security numbers.

The seemingly insurmountable difficulty of protecting this kind of data suggests that we are ready for a re-thinking of how we store important data. Obviously the encryption/password model isn’t working. Or at least the current implementations aren’t.

DANGER! Marketer Trying to Design A Product

Perhaps some the additional metadata fields being added to advanced file systems could be used for a OS-based encryption engine. Save a new file and the dialog box asks, in addition to the usual stuff, if the file should be encrypted and who (owner, group) should be allowed to decrypt it.

Perhaps an admin level account could require that all files going out USB ports be encrypted. Or ??

This is a real problem. I’m not technical enough to design a solution, but it seems like the current processes are hopelessly broken. Any creative engineers out there with some ideas? This could be a very popular utility.

Well, after getting /.’d yesterday this is bound to be an anti-climax. But I scoured the exhibition show floor the coolest products at SNW last night and found three contenders.

The first is Wasabi Systems the maker of a small (~1″x 3″) flash drive loaded with software that plugs into a motherboard’s IDE port and turns an ordinary pizza box server into either a NAS device, an IP-SAN device or both, with a software RAID module as an option. This all runs on Wasabi’s proprietary version of BSD, called Wasabi Certified BSD (please, let’s not get into a flame war about BSD vs Linux or GPL vs BSD licensing - I’m not endorsing or dissing either — just reporting). What’s cool is that in a few minutes a standard server (get Wasabi’s support list here) can become a NAS or iSCSI device. This is a natural for the SMB market where cost-effectiveness, ease of use and implementation speed are critical. It looks like DataCore’s SanSymphony product done right, i.e. on a stable Unix base.

Next on the list a company called Index Engines that “indexes data at wire speeds” or at “up to” 2Gbit/s all the data from your backup software (NetBackup, NetWorker, Tivoli Storage Manager) as it goes off to disk or tape. The indices are about 8% the size of the original data, so a server with 1TB of storage can index about 12TB of data. They claim to do full content indexing of more than 80 different document types, including pdf, doc, xls, pst and Exchange. Plus you can cluster up to 64 Index Engines together “providing unified search results for over 4 billion documents” (do I sense a 32-bit document address space here?). You access a Google-like interface through a browser to search.

While I am not crazy about the apparent requirement to index the data as it is backed up, I do believe this concept will help hasten the long overdue death of hierarchical file systems and the folder metaphor they support. Search and metadata extensions (such as IE’s indices) aren’t just the coming thing, they are the only thing, IMHO.

And finally, partly on the basis of the best and most extensible elevator pitch I’ve heard in a long time, Availl’s Wide Area File System. They claim that “No matter where users are, or how many open the same file at the same time, Availl ensures that only one user has read/write control.” Sounds like Einstein’s “spooky action at a distance” take on entangled quantum particles, but they stress that their byte-level differencing technology is what keeps bandwidth requirements to a minimum and latency low. All data is stored locally, presumably only the lock/unlock commands and byte differences get communicated, so data access is very fast. They mirror all the data in the WAFS across all the sites, so users can operate even when the network is down.

As to which is the coolest — well, you can decide for yourself. I like the Wasabi product best and plan to learn more about it. My biggest question: if it is as easy to use and configure as they say, why not sell a version directly to end-users to plug into already owned servers?

The skirmish between rotating and solid state storage (SSS) continues. Solid state storage proponents have lost a number of bets over the years to the rotating media barons. The bet? When solid state storage would overtake rotating storage in the market place.

With the emergence of solid-state USB flash drives (”pen” drives), it looked like SSS was mounting a credible attack on disk storage at the very low end. And sure enough, Iomega has taken it in the shorts as small, cheap and reliable pen drives have wiped out the market for Zip drives. Now I am seeing 1GB pen drives for $70, an incredible reduction in price from the $1 per MB price of just four years ago for the first pen drives.

But rotating storage forces have staged a counter-attack that looks to bound the pen drive market. Fry’s is offering a 2.2GB USB mini-drive for $89, or about $40/GB, almost half of what flash has achieved after four years of effort and high volume sales. It is too early to say if the new mini-drives will take off, but it a clear shot across the bow of SSS forces. While I love the small size and convenience of pen drives, history suggests that within a year we will see 8-10GB USB mini-drives for under $150.

It looks like pen drives will be confined to a sub-$100 niche and are not the beginning of the eventual take-over by SSS of higher price point storage markets.

SolarPC has announced the SolarLite PC, a $100 (when you buy $10M worth!) PC based on the VIA chipset and DSLinux (Damn Small Linux) which includes a fairly complete set of tiny apps in just 50MB. It isn’t clear if it includes a display in that price — there is a mention of a 7″ LCD screen — but even that size would push the price limit. Runs on a 12 volt battery. Check it out at http://linuxpr.com/releases/7357.html

And guess what: no hard drive! There is a compact flash drive, in keeping with their admirable low power consumption and no moving parts philosophy. For that price they might be able to affort a 128MB CF drive. Not a word about performance, which will stink given the modest bandwidth and slow access times of standard CF. Yet compared to the original dual-floppy PCs or the single floppy 128k Mac of 20 years ago, it will scream. It will do everything that people need doing, for a low price. What could be better?

Pardon my parochial interest, but some large scale storage would be nice. Computing in the now is nice, but persistence is the core of memory, learning and culture. OK. so let’s not freight a cheap computer with too much cultural baggage. But what is the point of a computer for the masses with out masses of storage?

This isn’t the fault of the SolarPC folks. Hard drives are too expensive. Not just dollars, but power, cooling, fragility and for the maximum price/performance, size. For the masses, what is the right, or minimum amount of mass storage? It depends. But 10 to 20x the size of the OS and basic apps (or 500 to 1,000 MB for DSLinux) seems like a fair start.

What about backup and data transfer? I suppose CF is reliable enough that backup is less important, but accidents still happen. And one could use the network. assuming there is one, for data transfer. But it is very helpful to be able to archive work, to move large files without a network, to publish, samizdat-style, without putting it up on the web for everyone to see.

Kudos to SolarPC for bringing a sensibly designed low-cost PC to market. I hope the next iteration will have more and better storage capabilities.

The latest from the Bad Ideas Never Die department: a number of vendors are postulating a really cheap PC that is, get this, network-based. Back in the dark ages of coal-fired mainframes, these things used to be called terminals, and they weren’t that popular even then. Since then every successful personal computing device has had significant persistent memory. So poor people are supposed to go without?

The basic idea seems to be that with today’s PC-on-a-chip systems a really cheap PC could be built.

The goofiness is in removing the hard drive to save money. Presumably users would store their data on the network. So where are all the millions of people who can’t afford a really stripped PC but can afford broadband to the home? The fact is that storage capacity is much cheaper than bandwidth from a total cost of ownership perspective.

If these guys really want to sell a billion really cheap PCs, they need to get a disk manufacturer to build a really cheap disk. It would be slow (3600 rpm or less) and optimized for low cost, maybe even using just one head, with smaller motors required and slower chips, small buffers, and none of the diagnostics and predictive failure that most drives have today. Put the whole disc controller on a single cheap chip. A 20GB $20 drive would make a $100 PC something people could use for years for websurfing, email, and basic office applications.

I might even buy one myself.