i've been waiting for YEARS for memory drives to take off..

i used to mess about with ramdrive.sys on the old DOS boxes..

HURRY UP!

i want at least reasonably priced 16gb ones for now..and cheaper bigger ones in the long run.

WE WANT SSD's!!!

Damn, that's a lot of solid state storage and a lot of bandwidth.

Now some OEM needs to put an array of those in a rack with a gigabit NIC as a NAS appliance and then I need to win the lottery several times in a row ...

Paris, because she also has expensive tastes.

No SAS? What year is this again?

Out of curiosity does anyone know the reliability of these things when compared to standard mechanical hard drives? Do you still need to build fancy RAID arrays with them, or is much of that sort of pallaver redundant ('scuse the pun)? Removing or simplifying RAID overheads could help to speed up throughput even further.

Of course strictly speaking the "I" in RAID would actually be something of a misnomer in these cases...

Because SCSI is not obsolete when it comes to enterprise gear, it is still used in the majority of high end storage arrays due to it's reliability and high speed. If I wanted to sell lots of these things, I would put it in a form factor where I could easily punt them to the big NAS and SAN storage companies, e.g. EMC, NetApp, HP, IBM, Sun. All of which use SCSI disks for their fastest storage.

Interestingly, a technology like this would significantly impact on the performance for Storage, at present a major limiting factor on NAS and SAN gear is the "seek time" or the time it takes to move the disk heads before a read or write, NetApp goes some way to get around this with their WAFL (Write Anywhere File Layout) technology, in which NetApp writes data to the closest free space, instead of moving the heads to the next contiguous space, NetApp then write a pointer to the data in a table called an inode table.

The traditional way to improve performance on storage is to use many smaller disks so that you increase the number of spindles and therefore your capacity to read and write sequentially.

In the case of SSD there are no disk heads to move, so the performance for other storage vendors will be massively improved while the benefit to NetApp will not be as great, it'll be interesting to see benchmarks when SSD's become used in the mainstream, especially how it affects the SAN vendors.

Also it may well be possible to gain very high performance from a smaller array, once the bottleneck of the disk heads is removed, although this will depend very much on contention introduced by the bus technology which is being used to connect the SSD's to the storage controller and the speed of the memory itself.

No there is no particular need for raid per se but you still want to carve these up into logical units to support whatever applications you're going to use them with, eg, multiple storage areas for database data, logs and so on.

Re OMFG, you can grab some now from Sun, yes you might want to store up a few days interest on your lottery million first.

http://uk.sun.com/products/complementary/pricebook/ssd4/

get <i>to</i> da choppa!

..i'll get me coat

used to share a house with a german student who did the greatest impression of that

PH because ppl like her about as much as they like ppl who correct quotes that eeeeeveryone is fully aware of :D

"Because SCSI is not obsolete when it comes to enterprise gear, it is still used in the majority of high end storage arrays due to it's reliability and high speed."

Do you have any idea what SAS is? Do you think before you post?

Actually, SCSI 320 is 320MegaBytes/sec (equivalent to ~2.4Gigabits/sec) compared to SAS which is running 1.6, 3.0, and 6.0 Gbps. There isn't that much difference between the 1.6 & 2.0Gbps and the SCSI 320.

The main advantage to SAS is the simplicity of the interface and cables (which makes the manufacturer more money.) Parallel SCSI is much more finicky.

For anyone who doesn't know what SAS is, i found this definition from google:

Serial Attached SCSI (SAS) is a new generation serial communication protocol for devices designed to allow for much higher speed data transfers and is compatible with SATA. ...

http://www.google.co.uk/search?hl=en&rlz=1B3GGGL_enGB235GB236&pwst=1&defl=en&q=define:SAS&sa=X&oi=glossary_definition&ct=title

Its an old article but no less educating http://www.digit-life.com/articles2/sas-and-sata/

Now, back to the plot. If SSDs are really going to match the speeds deliverable over SAS/SATA or whatever else then I want some of that. The time for running the traditional HDD has to be drawing to a close, especially so if they can get the price/performance to match what is around today. Early HDDs used to cost a fortune (I recall £100/MB just over 20 years ago), SDDs are going the same way, only the scaling will be quicker - you won't have to wait 20 years.

This is all good, very good

The wide cables are more of a liability with mechanical drives, with all that vibration and hot swapping. It probably makes sense to use up stocks of wide SCSI kit in this way. As a proof-of-concept for a single board solution this is how I would prototype it myself.

It's possible that military apps use older tech, such as SCSI, because of the lead time om procurement, and the expected life of the hardware.

Ever thought of what the recoil from firing a 5-inch gun requires in shockproofing computer systems on warship? There could be a nice niche market in replacement of spinning-disk systems in military applications.

..."equivalent to a seven and a half foot drop"? And would you guys like to explain how a drop is equivalent to 1500G? It's not the drop, it's the STOP!

>>> 1,500 Gs of shock, the equivalent of a seven-and-a-half foot drop<<<

The acceleration some critter receives when it finishes a drop is related to its velocity near the end of the drop (which is related to how far it falls, as you point out) but also has quite a bit to do with how quickly it stops, and that has quite a bit to do with how the critter is wrapped and what it falls on.

Something wrapped in foam is not going to see anywhere near as many Gs as something bolted into the insides of a laptop. May we presume the 1,500 Gs is what the laptop bay of a ruggedized Toshiba (or equivalent) sees after a 7.5 foot drop onto... concrete? Marshmallows? Another laptop that's also in the elevator with it and that has - and here's the tricky part - already bounced off the bottom of the elevator which just fell 7.5 feet before catching itself?

You're correct.

Unlike commercial IT users, military can't just go pick up whichever brand new high piece of kit happens to be on the market this week. The procurement, development, testing etc processes take a long time.

What about SSD failure. I can see a scenario where if an SSD fails you will lose 100% of your data since the failure is likely to be catastrophic. As we know with HDD failures, it is likely that most (if not all) of the data may be recoverable using well established techniques, if there is a need to do so.

Just to clear the air a bit here on low capacity SSD. There is a very simple and effective solution.

CompactFlash.

The latest generation of CF cards can run 40 MegaBytes (320 mbit/s) data transfer rates. The CF cards fully support the PATA data standard with UDMA. All that is required is an adapter to 40 pin ide for desktops or 44pin ide for notebooks. The computer recognises the flash card as a standard disk and no special drivers are required.

It is probably not a good idea to run swap on the CF, but otherwise the implementation is identical to what is installed in the Asus eeePC and other SSD systems released recently.

Acceleration also has a link to the relative density of the thing it is in when dropping. Terminal Velocity of a less dense 'thing' is reached very quickly and damage won't be as great, where as the more dense things will be accelerating when it hits the floor. (My physics tutor explained the flea, cat, man and elephant chucked from blackpool tower. Flea lives, cat has broken leg, man dies, elephant becomes pizza) So a server rack of these things falling seven feet will do a lot more damage than an asus eee with one of the things.

...but either way I want a 1 TB SSD for my asus please. And without a euromillions win first. :)

*Pedantry.

.....Corner Alert.....Pedants Corner Alert.......Pedants Corner Alert.......Pedants Cor...

The example given isn't so much a density thing as a scale thing.

Mass increases with the volume (that is to say (extent measurement)^3) whereas surface area goes with (extent measurement)^2.

Terminal velocity = sqrt(mg/<rho>AC) (where m=mass; g=accln. due to grav.; <rho> is the density of the medium; A is the presented area and C is the coeff. of friction)

All other things being equal, therefore, the terminal velocity increases with sqrt(r), as (m/A) is proportional to r^3/r^2 = r....

Got lots of SCSI2 and 3, and a couple of SATAs, no SAS.

Therefore SCSI is useful, SAS is not.

Seems like raid would be nice, as I understand that the SSDs have severely lower seek times than standard drives. Thus striping (probably with some parity blocks) would have some excellent performance characteristics. Why take 230MB/s when you can have 3 times that?