Connect 2MB with IDE ??

[Tech Support for IDE-CF]

Thanks for this post.
I dont know how to wire all this.
Do you think that in theory i could connect more than 3 motherboard with one
of this controller, (mb and master and slave)?

You would need one controller per motherboard. Dual ports memory can
connect them together. You can have a single master serving many
slave. The rest is just software.

Do you think that this kind of product will be made soon, many people are
connecting old computers together ...

That's just theorical design. There are many considerations in
developing such a product. However, it would be a good research
project for some grad students.

 

[RosalieM]

You can get cheap 4M Compact Flash and yank out the controller.

You would need one controller per motherboard. Dual ports memory can
connect them together. You can have a single master serving many
slave. The rest is just software.

So, i connect mb ide to this controller, and then each controllers together
?
But i need what kind of connection physically for controller to controller?
I know C language, i have book about linux drivers and linux kernel, i look
into kernel sources but i am far from understanding all, is this knowledge
enougth to begin?
Thanks for all.

 

[RosalieM]

So, i connect mb ide to this controller, and then each controllers together
?
But i need what kind of connection physically for controller to controller?
I know C language, i have book about linux drivers and linux kernel, i look
into kernel sources but i am far from understanding all, is this knowledge
enougth to begin?
Thanks for all.

did i wrote something very stupid ?
i am still interested in doing this.

 

[kony]

did i wrote something very stupid ?
i am still interested in doing this.

This solution will be slow, costly, relatively large, and IF you got it
working then all you'd have is a limited-write-cycle shared drive.

If THAT is all you wanted, you're still better off to just use low-height
PCI network cards. They'd be as fast, less expensive, and meant for
exactly this use so the system(s) are up and working right away, not so
many months later than it'd be more time and cost effective to just buy a
new motherboard and CPU.

 

[RosalieM]

What better experience to learn ? I think it should be faster anyway.

Buy new cards, new soft, dont be interested how stuff works, patch them if
pb, patch the patch...Then try new hardware, new soft, new patches, patches
for new patches ... Again

If it is possible i would like to try something else. I dont beleive i would
succeed, but i am confident that someone would make such ideas succeed. When
i discovered python i felt confident for that.

And as someone told me here that it may be possible ...

 

[Tech Support for IDE-CF]

So, i connect mb ide to this controller, and then each controllers together
?
But i need what kind of connection physically for controller to controller?

It depends on the controller. For instance, the controller in our 4M
Compact Flash Drive handles multiplexed address/data in approx. 10
signal lines. The 20M & 32M controolers use separate signal lines
(approx 30 lines). In that case, you can just wire them directly to a
dual ports ram chip.

 

[RosalieM]

It depends on the controller. For instance, the controller in our 4M

Compact Flash Drive handles multiplexed address/data in approx. 10
signal lines. The 20M & 32M controolers use separate signal lines
(approx 30 lines). In that case, you can just wire them directly to a
dual ports ram chip.

Thanks, in fact i need soldering iron or something like that? Ok i must give
up.
Thanks for your answers.

 

[*Vanguard*]

"kony" said in news:[email protected]:

This solution will be slow, costly, relatively large, and IF you got
it working then all you'd have is a limited-write-cycle shared drive.

If THAT is all you wanted, you're still better off to just use
low-height PCI network cards. They'd be as fast, less expensive, and
meant for exactly this use so the system(s) are up and working right
away, not so many months later than it'd be more time and cost
effective to just buy a new motherboard and CPU.

Yeah, and with everything onboard (video, network controllers, IDE or
SATA, sound, and USB) you could stack the mobos very close and use the
network controllers with a switch hub to connect them all. Some onboard
network controllers can go to 100 GB so you have the 75% of the PCI's
bandwidth. You'd have the solution NOW plus you would be using standard
protocols to communicate rather than have to make your own proprietary
ones.

In fact, seems like there are solutions already like this where you get
rack mounted computers and switches. Get a rack-mounted case, like
http://www.gtweb.net/j1100.html (there are smaller units that don't have
the 3.5" and 5.25" ext. drive bays if you don't need a floppy and CD/DVD
drive), to put in your motherboard with the onboard video and Gigabyte
LAN, get a rack-mounted switch, probably a large enough UPS to handle
all the power (rack-mounted optional), the rack, of course, and you're
off with a custom-built multiple system computing center. However, rack
systems are always expensive.

 

[Tech Support for IDE-CF]

This solution will be slow,

Approx. 2M bytes / seconds, vs. approx 800 K bytes / seconds on USB 1
(most older MB has USB 1 only.

costly,

The CF controllers on 4M CF are $1 each.

relatively large,

Approx. 3" x 1" pcb should do it.

and IF you got it working

That's the hard part.

then all you'd have is a limited-write-cycle shared drive.

Not if you use static ram instead of flash.

 

[kony]

What better experience to learn ? I think it should be faster anyway.

Have you studied any specs for ATA? That would be prerequisite learning.
You'd see that there is no point to going though an IDE channel to do
something other than hooking up a drive.

Buy new cards, new soft, dont be interested how stuff works, patch them if
pb, patch the patch...Then try new hardware, new soft, new patches, patches
for new patches ... Again

Yes, buy parts engineered to attain the goal. There's a vast difference
between being interested in how stuff works and trying to reinvent the
wheel though a very narrow focus that has no possibility other than an
inferior end-result. If the end result is more costly and lower
performance, usability than a ready-made solution then it's quite clear
that the project has no merit. The goal must be worthwhile.

If it is possible i would like to try something else. I dont beleive i would
succeed, but i am confident that someone would make such ideas succeed. When
i discovered python i felt confident for that.

And as someone told me here that it may be possible ...

Possible to share a flash drive? Yes, though a LOT of work you could do
as I already mentioned, attain a slow limited write cycle shared drive
that would be large enough to offset any space savings seen by trying to
stack boards together as close as possible.

If you want to learn about IDE controllers or clustering systems that's
great, but the IDE controller isn't the way to do it.

 

[kony]

Approx. 2M bytes / seconds, vs. approx 800 K bytes / seconds on USB 1
(most older MB has USB 1 only.

I wouldn't recommend USB1 either. PCI interface card, either proprietary
or standard NIC.

The CF controllers on 4M CF are $1 each.

They're in a ready to assemble form? Seems like it'll take a lot of $ to
buy the precision tools too. Then again can anyone just spend $5 and have
5 CF controllers? Costs rise, you can't only count a raw part cost for
one component.

Let's look ahead a bit... PCBs, wiring, tools, the memory itself,
enclosures, relays, etc, not to mention the amount of time spend in this.
A little bit of time on a hobby is great, but a LOT of time is worth $.

Not if you use static ram instead of flash.

.... and the cost keeps rising, or has SRAM plummeted in price recently?

 

[RosalieM]

I wouldn't recommend USB1 either. PCI interface card, either proprietary
or standard NIC.

Hi Kony.
Thanks for ATA programming, i will read about this.
You are right, complications to wire are far from the scope of what i am
able to do.
About usb1 if you read the post yes it is old motherboard paid 5$ each.
I wanted to learn futher in driver writing in fact.

I asked to know how to make ide communication, and believe that i understand
better what it means. Thanks all.