Slashdot Mirror

What is the expected lifespan (in cycles) for flash memory? I thought it was only good for a few thousand writes.

As I understand it, this is basically a solved problem. For one thing, the number of write cycles before a location dies is actually more like 10,000 or 100,000.

But more importantly, apparently most all modern flash controllers automatically and transparently cycle writes through various parts of the flash, so that if you tell the controller to write to the same block several times in a row, behind the scenes it will write to different physical blocks in the flash each time. It somehow maintains a constantly-changing mapping between block addresses and actual, physical blocks.

This process is called "wear leveling", and by doing this, the problem of wearing out one particular location is mostly eliminated. In fact, given the number of cycles that each block can be written, and given the transfer rate of a flash device, it might take years and years of continuous writing 24 hours a day before you can wear it out. Eventually, it will wear out, but most hard drives wouldn't stand up well to that kind of activity either, so you're no worse off with flash.

This article has a section on wear-leveling.

I do have to agree that citing a buzz-phrase like digital vaults is a very lame way to end an artical. But that said, it's an interesting technology to apply to solve this problem.

Mind you, it doesn't actually solve it as many unsolved issues remain (escalation of privledges from within the application, administartive access, development backdoors, key management, migration to the new architecture, etc.) but it's nice to have a new tool to apply to the problem.

Below is an explainatin of "Digital Vaults":

Digital Vaults enable users across the internet to share access to sensitive information in a simple secure way.

A major challenge that is faced by all organisations selecting IT technology is trying to clearly understand how a particular solution may address the challenges they are tasked with solving. And this often involves trying to understand what various vendors mean when using generic terminology.

The term "Digital Vault" has come to the fore in the last few months and now several vendors are offering technology under the umbrella of digital vaulting. So what should you understand? A simple acid test to apply to anything claiming to be a digital vault is the following. Does the digital vault hide items from those who have no right to see them, and does it ensure that those with access rights are monitored every step of the way.

The term vault should be used because it relates to the vault in the physical world. Every enterprise relies on few priceless items that must never be lost or exposed. The danger of losing or exposing these priceless items is vital to the enterprise's business continuity and can even threaten its very existence. In today's business world, a large percentage of those items is in digital format. Most business enterprises today will still use the physical vault to securely store copies of the critical data, but this is impractical when on the one hand you are required to make that data available on a day to day basis for those who need to view, and modify the data, and at the same time you are required to keep it under "lock and key" so that those who are not entitled to see it are kept away from it.

Bringing it back to the physical world analogy; the physical vault can only be accessed by those who have privileges to do so, and once in the vault, only those safety deposit boxes that you have the right to open should be made available to you. For those who saw the the Bourne Identity (movie), you may remember the scene when the hero enters the bank and gains access to the vault. He is then provided access to his private safety deposit box - well the digital vault needs to mirror this physical scenario. So the digital vault should be a mirror image of the physical vault. Critical data needs to be stored in a secure location, and should be visible only to those with the rights to see it.

Another key factor in identifying a Digital Vault should be its ability to mimic all existing security processes and procedures in the organisation for handling sensitive information. For example, most organisations will have clearly defined policies and procedures defining how sensitive physical items are handled. For example, who has access to the physical vault, and the security boxes? Are individuals allowed to access on their own, or is a dual control mechanism in place, for example dual keys? Does staff have to be authorized to enter, and are there times of day when access is permitted. These and many more procedures are found in organisations, and a Digital Vault must be able to address these procedures as is. It is not advisable to try and redefine policies and procedures to fit technology - the technology has to fit.

A digital vault by its very nature is going to provide some standard services to ensure that its contents are protected, such as being a long-term repository, highly secured regardless of overall network security and regardless of the physical topology of th

I had one of those Cheap and Crappy (TM) Western Digital USB drives fail on me after THIRTEEN months. Guess how long the warranty was: TWELVE months! After being casually rebuffed by customer support I vowed never to buy another WD drive (even an internal one).

I go with SCSI now. SCSI boot drives are amazing. My C: drive is a 18 GB, Seagate Cheetah (15krpm) that's four years old and my D: drive is a 72 GB IBM/Hitachi 10krpm. I use both extensively for intensive applications (like video capture, encoding) and have never had a problem with SCSI so far (though you pay a premium for it).

Personally I wish SSD could be made for the masses, but sadly they would probably suffer from the same quality issues that cheap platter drives do nowadays. Just like everything that's commoditized. Anyway, check out http://www.storagesearch.com/ssd-buyers-guide.html if you're looking for some info on SSD. Note that prices are not included because no manufacturer wants you to know how bloody expensive these things are until they can real you in with an email or telephone call.

You will likely be better served if you just use a solid state disk for this.

IDE and SATA != spin disk magnetic storage.
Different technologies already exist, like CD-ROM :) or even stranger like flash drives or solid state drives (with zoooooming performances).

I think this link can be usefull if you're interessed in this technologies:

http://www.storagesearch.com/ssd-buyers-guide.html

At 6.25Gb/s, about 6 seconds. I can hear MPAA quaking in their boots. Or, if you prefer, you could stream about 600 DVDs simultaneously. *drool*

Alas, even RAM-based SAN devices can't keep up with that bandwidth by half. Time to use latency of network loops as a storage mechanism. =)

" ... With the amounts of RAM and flash memory available these days, I don't see why we couldn't run an OS off one. ..."

Well, you can. For information on extremely large (and expensive) solid state disks, see storagesearch.com.
For smaller needs (desktop and up) you can use RAM drives from platypus technologies who offer PCI card (and other) solutions.

Platypus PCI cards (QuikDRIVE) offer up to 8GB per slot (can be combined to form larger virtual drives) for about $2 to 10K per card. They are designed to read from and back up to regular HDs upon startup/shutdown. Essentially you can run the OS and all applications in RAM (the platypus drives are seen by the hardware as the startup drives) along with whatever documents fit. The OS sees the cards as hard drives, so as far as the user is concerned, they act as if they were.

This type of application of solid state storage requires higher spec RAM than "regular" sticks, so it's a bit pricy, but hardly out of the question. A lot of companies are using them now (eg Monster.com). Check out Section 6.0 Applications on the Platypus site for more info on currently deployed uses. Mac, Linux, UNIX & Windows compatible.