Hard Drive and other Disk Interfaces

[Arcadian]

I wanted to spawn a discussion of hard drive and other disk interfaces. Let's talk about the present and future states of these interfaces, and how they relate to the desktop and server markets.

On the desktop side, we have the old trusty IDE interface. Do you believe this interface is running out of steam, or is there enough to keep it going for a few more years? Keep in mind that current computer needs don't even surpass ATA/66, let alone the new ATA/100 interface. Do we even need to continue to upgrade this technology? Also, what do people feel about the emergence of IEEE 1394 (firewire) technology in disk drives, or the Serial ATA interface previewed at Intel's Developer Forum last month? Do you feel these technologies can take over, and if so, how?

In terms of the server world, we have the old and trusty SCSI standard, which is still used extensively in high performance RAID arrays. Do you still see this technologyg remaining popular as it moves to faster and faster speeds? I have heard of 320SCSI and 640SCSI as being technologies on the horizon that can double and quadruple performance. Also, how about Fibre Channel? Does anybody out there have experience with this standard, and do they feel it will continue in the future? There is also talk of IEEE 1394 finding its way into the server world, as well as Intel's Infiniband technology, which works like Fibre Channel, but is much more highly scalable.

Let's try to answer some of these questions with some hard core technical speculation! If anyone has any links to more information, feel free to post. Even if you want to offer opinions, bring it on... this could be an interesting discussion. http://www.sharkyforums.com/ubb/smile.gif

[zombor]

firewire wont catch on as a disk interface (in desktops anyway) the same reason SCSI didnt: price. I saw in a mac ad magazine a 30Gb firewire HD for oh...around $1000. Then the next page, a IDE 72Gb was around $450(??). That extra speed increase isnt nearly enuf to warrent the price. In server, it could be a different story, becuase price usually isnt an issue.

As for IDE, "why fix it if it aint broken?" lives up to its name. In order to utilize the full potential of any of our components we need to start usinga new architecture that is designed to be used at these speeds, and can handle these speeds. *** is up with the system bus not being able to handle the memory bus on the PC mobo? This needs to be fixed in parrallel to the disk drive issue, if you know what i mean. In PCs ide is more than sufficent to use.


------------------
Compaq Armada E500
650MHz PIII 128MB RAM
8MB ATI rage mobility
11GB hard drive/DVD
dual booting win98/2k

who said you cant game on a notebook????

[This message has been edited by zombor (edited October 06, 2000).]

[jtshaw]

Honestly, I don't see Firewire taking the place of IDE anytime soon. It is really, really expensive. I was looking into using it in something I was designing and it turned out to be over 6x the price of what I ended up using. Give I wasn't doing anything with disk drives it was still really expensive.

On the desktop level I think we will see IDE for much time to come. Currently the ATA66 and 100 standards aren't even being used to there full capacity. I have found most drives that say they are ATA100 or ATA66 compatible mean they can interface over the cable, not that they sustain the full transfer rate of the spec.

On the server level SCSI (the small computer system interface....badly misnamed:P) RAIDs (something about an array of INEXPENSIVE?!? disks...I guess inexpensive is all relative eh?) will still probably lead the pack for a while, although I suppose we could see firewire sneak in. I don't see fiberchannel in the near future because that is also really expensive, although if it is to appear anywhere I would guess it would be in very high end servers. I don't know that going to something like fiber channel is really that neccesary. SCSI160 is damn fast and if SCSI320 or SCSI640 were to show up that seams like it would do the trick for a little while.

Although the above is what I would guess will happen I am a huge fan of well overpowered hardware myself http://www.sharkyforums.com/ubb/smile.gif I would love to see ieee1394 and fiberchannel (or Intels version) make it into computers that I could go spend all my hard earn cash on http://www.sharkyforums.com/ubb/smile.gif

[SpikeyFreak]

IDE is going to be around for a long time. Especially if more companies start putting more than 2 controllers on the mobo like the Abit RAID boards.

FireWire is expensive now, but do you think it will always be? Look at RDRAM prices. Hopefully more people will start making IEEE 1394, and therefore drive the price down.

RAID, according to MS, stands for Redundant Array of Inexpisive Disks. According to other people though, it's Independant Disks (which makes a lot more sense).

I don't see SCSI becoming a home PC thing. It's too expensive. Plus, you can get bigger IDE drives, and there are now IDE RAID solutions, so you can have security at home too.

--Carnal Spikey

[KR02]

I think that Firewire will stay where it is: high end vidio connections. As for Scisi, i don't think that it'll ever become a HomePC thing. I see ATA staying for a while, but the DMA levels will contue to move up. I do, however, think that the Hard Drives themselves will change. I tihnk they will change by being made out of different materials and haveing more heads and a faster rotational speed. That way they will be able to keep up with the ATA speeds. Same will happen for SCSI, but i still don't think that it will keep up.

------------------
Why Drink and Drive when you can Smoke and Fly?

[phantom]

i'm holding out for holographic molecular memory
when that comes around, we won't need hard drives anymore
or any drives/cards for that matter
everything will be run from on-board ram with no moving parts
and data throughput will occur in logrithmic boulean logic streams from each molecular data point (up, down, right, left, right, back, forth, diagonals)
the limiting factor to this new nanotech
will be the physiography of the human interface
so save those notebook platforms - they'll be the mainframes of tomorrow

------------------
phantom ... out.

[Warin]

Originally posted by phantom:
i'm holding out for holographic molecular memory
when that comes around, we won't need hard drives anymore......


Not to be too negative, but this is an example of the worst kind of 'Pie in the Sky' sort of thinking. There is more fiction than science in these concepts, many of which havent been proved in any labs...and if they have, I'd like to see the information!

SCSI is superior to IDE, in that it offers more flexibility in the number of connected devices and in the speed/processor loading. It is more expensive, and sadly lower SCSI prices are a chicken/egg situation. More consumers will purchase cheap SCSI..but cheap SCSI wont happen until more consumers buy in...

As for the name...

SCSI was developed when 640k WAS enough, and Small meant 'Smaller than a Mainframe'

But calling it 'Scuzzy' was justa BAD idea!

------------------
No Matter Where You Go, There You Are.

[This message has been edited by Warin (edited October 08, 2000).]

[Warin]

Originally posted by Warin:
Not to be too negative, but this is an example of the worst kind of 'Pie in the Sky' sort of thinking. There is more fiction than science in these concepts, many of which havent been proved in any labs...and if they have, I'd like to see the information!

SCSI is superior to IDE, in that it offers more flexibility in the number of connected devices, and in the speed/processor loading. It is more expensive, and sadly lower SCSI prices are a chicken/egg situation. More consumers will purchase cheap SCSI..but cheap SCSI wont happen until more consumers buy in...

As for the name...

SCSI was developed when 640k WAS enough, and Small meant 'Smaller than a Mainframe'

But calling it 'Scuzzy' was justa BAD idea!

[Warin]

Gack!

I suck! it's EDIT, not reply!!!
Damnit!


------------------
No Matter Where You Go, There You Are.

[phantom]

today's science fiction is tomorrow's reality, warin.
besides, arcadian's original post said: "Let's try to answer some of these questions with some hard core technical speculation!"
operative terms - hard core technical speculation.
i think it would be cool to have everything ever written in the entire history of mankind
on a memory chip the size of a sugar cube
and still have over half of the ram capacity left over for user interfaces
consider the benefits: no moving parts, no significant heat issues, molecular transistors, and what not
for the last four years, the aberdeen research lab and jpl
have been slowly working out the theoretical bugs on holographic molecular memory
for obvious reasons, the military is quite interested in this tech


------------------
phantom ... out.

[Arcadian]

Originally posted by phantom:
today's science fiction is tomorrow's reality, warin.
besides, arcadian's original post said: "Let's try to answer some of these questions with some hard core technical speculation!"
operative terms - hard core technical speculation.
i think it would be cool to have everything ever written in the entire history of mankind
on a memory chip the size of a sugar cube
and still have over half of the ram capacity left over for user interfaces
consider the benefits: no moving parts, no significant heat issues, molecular transistors, and what not
for the last four years, the aberdeen research lab and jpl
have been slowly working out the theoretical bugs on holographic molecular memory
for obvious reasons, the military is quite interested in this tech


Wow... I guess I expect JPL to be the leading researchers for the millitary. That's all very interesting long term speculation, phantom. I was also wondering what opinions you had for the short term, especially with some of the new technology that is actually starting to get implemented, like Firewire and Infiniband. I only ask because there seems to be a lot more information about these technologies as opposed to holographic molecular memory, though if you have any links to information on the latter, then by all means respond with those. I was hoping someone might have some insights to the pros and cons of upcoming technology that we might see next year, which would ignite some discussions regarding those. But I still enjoy long term speculation as well. http://www.sharkyforums.com/ubb/smile.gif

[Warin]

Hey! Dont get me wrong, Phantom!

I'd love to see these technologies! But the fact is that they dont yet exist in a meaningful manner! I find the best 'theoretical' computer sciences often come out of IBM's labs. Most of the high density hard drives of today can be traced back to work done there. I remember a Byte magazine 'round about 85 with an article on a new head technology being developed at IBM that might alow for hundreds and hundreds of megabytes in a drive no larger than a paperback book! Wow!

Back to the question at hand...

I read a bit on serial ATA. I think that of all the current 'new' tech interfaces, this is the most promising. It uses a much simpler cable design that is faster, and can be longer than current cables...plus gets us past the four device limit on current IDE bus controllers.

I need to do some scans from another Byte I recently found..and let us all compare prices from then to now. For those that were very young in those days, it would open your eyes to how amazingly cheap the technology has become!



------------------
No Matter Where You Go, There You Are.

[jcortel]

I am getting more interested in RAID 0 compatible motherboards. Almost twice the speed and you lose none of size of the two drives, assuming they are the same size.

Abit makes the a BX and Via chipset RAID supporting motherboards. They have benchmarks for the RAID 0 setting.
http://hardocp.com/reviews/mainboards/abit/bx133/

[busithoth]

You know, I think IDE will hold it's ground for another 5 years at least. It really depends on how long the world's willing to invest large amounts of cash in developing newer, better and faster tech, even if it really surpasses the needs of 95% of the demand market. I was thinking about getting a SCSI drive, but raid changed my mind. Now I could just sit back and enjoy the IBM deskstar that I have, which seems fast enough for my needs, or I could get a RAID controller and do that funky thing.

IDE is so cheap right now, I think it's cementing itself even further as the real standard in the computing world.

Personally, I'd love to see Firewire get integrated into the structure of PC's, connecting the devices within, allowing amazing bandwidth potential well past 5 years from now. The cost would come down if it were more common. Nothing in the Mac world is going to be cheap. You pay through the nose for the benefit of not knowing crap about your computer and still having it run.

The fact that those peak transfer rates of 40-50MB/s on drives now (burst cache only, mind you) makes me rest assured that until there's a real demand for it, IDE will sit where it is-in the majority of systems, with the best price/performance option out there.

[Arcadian]

I happen to agree with a lot of people here that hard drive interfaces really don't need that much of an improvement on a desktop system. Most desktops don't have any more than 2 drives, so the interface isn't being stressed at all, and the drives themselves can't spin fast enough to create any decent transfer rates. Perhaps RAID 0 ATA cards may be able to breath more life into desktop systems, but they only seem to be a niche market right now.

As for server systems or especially storage farms, I definately see new technology necessary. When you have a system with over a hundred drives, the interface becomes the major bottleneck, which is why high speed PCI RAID cards (on the 66MHz 64bit PCI bus) are so popular in large enterprise servers. Perhaps PCI-X will offer the next path in I/O, but there will need to be even better hard drive interfaces to take advantage of that new bandwidth. Going wider used a be a good solution (wide SCSI - 16bit), but I'm starting to think serial may be the future, which is why I am so into Infiniband right now. If you haven't heard of Infiniband, you should definately look it up, because it is looking like a great new technology for large storage systems.