Hard Drive Primary/Secondary Channel Bandwidth

[d3vil]

I've always read that having 2 hard drives on 1 channel is a bad thing as it will slow down access due to both fighting for bandwidth on one channel, but is that all there is to it? Couldn't you just solve that problem by having more bandwidth for each channel?

If I had an ATA133 card/cable hooking up both hard drives would that be enough bandwidth to support 2 slow (E.g. ata66) hard drives? Being that 66 MB/sec x 2 <= 133 MB/sec

If that wouldn't work, then what would?

Thanks
Alex

 

[Cogman]

Well, from what I heard, And this was a long time ago so I dont know if it still holds true, You should not Use 2 IDE devices on the same cable that are diffreing Speeds, The reasion is that the IDE controler will automatically set the speed to the slower device choking the faster one. Now if they Are the same speed, I dont think you would really have much of a problem. There are really very few things that require Accessing both hds at the same time, hence I dont think you would see a drop in speed.

Someone Verify/Clarify what I just said?

 

[d3vil]

reading another forum post a while back, i've read that that is a myth..
the guy said there were tons of myth links about it, but i could only find one easily.

googling for "hard drive myth channel" i found this..
http://www.geocities.com/sheppola/hard.html

once again.. if having 2 ata66 drives doesn't set the whole channel to ata66, and it remains at ata133.. would it still be acceptable to have both on 1 channel?

 

[cquark]

Originally posted by: d3vil
reading another forum post a while back, i've read that that is a myth..
the guy said there were tons of myth links about it, but i could only find one easily.

googling for "hard drive myth channel" i found this..
http://www.geocities.com/sheppola/hard.html

once again.. if having 2 ata66 drives doesn't set the whole channel to ata66, and it remains at ata133.. would it still be acceptable to have both on 1 channel?

You'd have sufficient bandwidth to not have any problems, and that is mostly true even if the disks are both ATA133 disks as the fastest ATA disks can only do around 70 MB/s under best conditions, but there may be other limitations than bandwidth.

 

[Matthias99]

Originally posted by: cquark
You'd have sufficient bandwidth to not have any problems, and that is mostly true even if the disks are both ATA133 disks as the fastest ATA disks can only do around 70 MB/s under best conditions, but there may be other limitations than bandwidth.

In particular, you can only talk to *one* device at a time on an IDE channel. Thus, if you copy from one hard drive to the other on the same channel, it halves the available bandwidth. With ATA133 this is probably not much of an issue (since most IDE hard drives top out at well under 66MB/sec., even for sequential transfers), but with the slower protocols, it can seriously constrain transfer rates if you're moving large amounts of data around. With older CD burners (before they had large data buffers onboard) running at ATA33, it could even cause it to fail if you had both the source drive and burner on the same channel, or if something accessed the second drive on the burner's channel while it was working.

It also tends to slow the drives down even more than just halving the bandwidth would suggest -- rather than one disk doing "read, read, read, read" and the other doing "write, write, write, write", you basically get a pattern like "read, stop, read, stop" and "write, stop, write, stop". If both drives are the same speed and on different channels, in theory you can transfer a large file in the amount of time it takes to read or write it (since the destination drive can write it at the same time the source drive is reading it). On the same channel, you can only have one drive going at a time, so it will take twice as long.

 

[d3vil]

is hdparm under linux a good benchmark util for hard drive speed testing?
I was thinking of running hdparm simultaneously on 2 hard drives on the same channel to see how much it would affect it, vs. running it separately.

I don't have the 2 hard drives set up yet, but i'd be willing to try if this is a good test.

I'm mainly curious about this topic because i am setting up a samba file server on an old pc.. going to be putting a pci ata133 card in there, and since i only have one, i'm gonna have to stuff 4 harddrives onto 1 ata133 card.

 

[cquark]

Originally posted by: d3vil
is hdparm under linux a good benchmark util for hard drive speed testing?
I was thinking of running hdparm simultaneously on 2 hard drives on the same channel to see how much it would affect it, vs. running it separately.

I don't have the 2 hard drives set up yet, but i'd be willing to try if this is a good test.

I'm mainly curious about this topic because i am setting up a samba file server on an old pc.. going to be putting a pci ata133 card in there, and since i only have one, i'm gonna have to stuff 4 harddrives onto 1 ata133 card.

While hdparm is a poor benchmark in that it doesn't model normal disk usage, it should work fine for what you want to test--whether your hard disk transfer rates are limited by channel contention.

 

[Concillian]

I have two things to say:

1) Matthias99 is entirely correct

2) There is a reason that ATA-RAID cards put each HD on an independent channel, it is for just ther reasons that Matthias99 outlined.

I feel you do not need to go through with your testing, but if you feel the need... go right ahead.

You think RAID card manufacturers would have 4 channels on an IDE RAID 5 card when 2 would work just as fine? I think they'd rather save the manufacturing costs. In fact i know they would.

 

[Peter]

Matthias is entirely correct - and here's the reason why.

A master/slave arrangement on an IDE channel _IS_ a single unit. One controller unit (in the master) controls access to both physical media. Now since IDE controllers cannot work on more than one command at any given time, there is no interface bandwidth sharing in a master/slave setup.

(Please mind, when I say "controller", I'm referring to the controller in the drive.)