This is not an actual problem, or is it? I have Mandrive installed on an IDE drive (master) formatted to RaiserFS. I have a spare drive on the same IDE channel (slave) formatted to FAT32 so the space can be used by Windows and Mandriva. I tried to transfer a large file (a back-up of my mandriva DVD image) from the Mandrive drive to the FAT32 drive. The transfer rate did not go above 7 MB/s. So the ETA was something like 20+ minutes. Is this because of reading/writing accross two different FS's? Or could it be down to DMA setting etc? Thanks for looking.
There are many things that can cause sluggish transfers between HDDs in the same system. The first thing I'd look at is the IDE chain layout; are both HDDs on the same IDE channel? That can severely impair the performance of drive-to-drive transfers. Lack of DMA can also slow transfers and indeed bog down the entire system. You can check and see if your system is using the correct level of DMA by using the hdparm command: Code: /sbin/hdparm -i /dev/hd* Feel free to post the results here if you'd like a second opinion.
Yes, they are indeed on the same channel. The Mandriva sits on the master and the spare space drive is set to slave. I will check DMA later and post... If DMA is ON and transfer is still slow, I might get a PCI IDE expansion card and connect the spare drive to it. I am planning to keep this particular set-up and I do not want to wait so long for data transfers.... Do you think that will help the transfer rate? Many thanks AT.
Yep, looks good! It's on the highest DMA mode supported. Looks like your bottleneck is the drive configuration; HDD's don't like sharing IDE channels. Do you at least have them on an 80-conductor ribbon cable, or is it a 40-conductor cable?
Well, I never actualy counted.... HAHAHAHAHAHA Seriously, I never checked. I am pretty sure I am using the ribbon which came with the mobo, so I am assuming it is just a standard 40.... I wonder how expensive will the 80 flavour be.... What do you think about adding the PCI card? Thanks for looking.
If your OS detects UDMA, then you are using an 80pin cable. If you were using a 40pin cable, then you would be running in PIO modes.
I will take your word for it..... Well, in that case, not much more to do.... Just have a coffee break when transfering..... Thanks guys!
Not quite. A 40-conductor cable will limit the transfer rate to ATA-33, but not necessarily the direct memory access mode. ATA-33 cables allow up to UDMA2. However, since sabashuali's system is in fact reporting back UDMA5, we can rest assured that it is indeed using an 80 conductor cable. So, with that in mind, we can be fairly sure that the bottleneck in sabashuali's rig is the fact that two HDDs are transferring data across a single IDE cable. If I was you, sabashuali, I'd consider buying an inexpensive PCI IDE controller and placing the secondary HDD on its own IDE channel on that controller. That would increase your bandwidth immensely and increase your systems capacity to handle more devices overall. P.S. - I have attached a picture to illustrate the difference between the appearance of a 40-conductor ribbon cable and an 80-conductor.
What can I say... Is vis guy somefink or wa' ? Thanks AT. I probably will get the the expansion card. When I get my new case...... In other words.... when I have the money. Good reply! Thanks for the info!
Hmmm, I'm not sure I completely buy that explanation just yet Only the other day, I was transfering 40GB of data from one hard drive to another. They were both running at ATA-100, and sharing the same IDE cable. The whole transfering process took about 30-40 mins. I was getting extremely fast transfer speeds. The only difference is I was transfer data between two ReiserFS formatting drives, as opposed to sabashuali's ReiserFS - FAT32 combination. Could it be that it just takes longer for Linux to write information onto a FAT32 drive?
I've never had any slow transfer rates using two drives on the one chain either. That is my current setup and it transfers pretty fast.
If you want to see a truly crippled HDD array, build a RAID out of drives which share IDE channels. Yes, sharing IDE chains does slow transfer rates substantially. FAT32 is also a ghastly filesystem, and the larger the volume the slower it is. So while FAT32 itself is certainly a big part of the problem, the IDE configuration is also contributing to the situation. As for Reiser, that has absolutely nothing to do with it. If both volumes were FAT, it would still be dog slow.
[ot] I resent that! I wish the transfer was as 'slow' as my dog.... Please check here - http://www.hardwareforums.com/gallery/showimage.php?i=392&c=3&userid=2551 hehe [/ot]
