RAID with partitions
I'm curious what Windows would do with the following setup. Keep in mind that I will be using software RAID. My Windows drive is a 120GB SATA drive. I will partition this drive into two letters, C (40GB) and D (80 GB). I will install two more drives, both 80GB IDE drives (WD800JB). I would like to create a software RAID array with both IDE drives and the D partition. Do you think this would even work, and if so, would the array run faster just with the two IDE drives?
I have heard that it works, but have no first hand experience.
I have no idea if it would be any faster, but if you used RAID 5 it might be safer.
However you will only be able to access D: from inside windows, which means you can't backup your OS a file on D:.
Well technically you could with Ghost 9, but then you would have no way of restoring it, unless Ghost 9's recovery disc reads windows software raid?
If you have the time and the inclination I would download HDTach and benchmark it and see for yourself what transfer rates, access times and cpu utilization you get.
I am very curious so if it works and you run benchmarks please post them!
<P ID="edit"><FONT SIZE=-1><EM>Edited by Codesmith on 02/05/05 05:41 PM.</EM></FONT></P>
I have an update to provide everyone. I have all the drives installed and I have created a RAID out of the two IDE drives only, so far. I have found that benchmarking with HDbench is not helpful, because it does not detect (or perhaps it ignores) the software RAID. It only benches individual drives, so I will be benchmarking with SiSoft and Winbench, and will get back to you.
I need to find out also, do you lose data when you convert a drive to a dynamic drive? If not I will convert my C drive and use the second partition as the third part of the RAID.
Well I'm not sure what's going on now! SiSoft crashes whenever I try to bench the RAID array (or the C drive for that matter), but it will run the other benchmarks fine. Winbench99 did work, and the results for the RAID array (w/o SATA) were better than just the SATA drive alone, but in truth they were much less than I thought they'd be. 22800 Marks for the SATA drive and 35600 marks for the RAID array. I ran the HE test. I'll run the business test next and post the results. Any suggestions on how to get a better score?
Well I added the SATA drive to the software RAID equation and my preliminary benchmarking results are positive: My Winbench 99 score went from 9900 to 11800. I will do more benchmarking tomorrow, right now I must sleep!
I could quite possibly have the most unlikely RAID setup ever conceived by mankind! 2 IDE drives and one SATA drive split into two partitions, all running from an old AT power supply in software RAID!
RAID 5 Right? How hard was it to modify Windows XP to enable it?
Please run HDTach as well. I like seeing lowe level MB/s numbers. Also I want to know what price you are paying in CPU utilizatoin and access time. Apparently it worth it since the benchmarks simulate real applications.
Also before you put anything important on it, please do some reliablity testing. Start Windows with one drive disconnected and see if you can just restart with it attached. Start with two drives disconnected and see if it kills the array. Swap IDE ports and see if there are any problems.
It would suck if all your data is lost if you move things arround and forget to plug in one or two drives. I am hoping it will let you recover once you correct the problem without having to rebuild.
When I get enough money for another WD740GB I will put the Raptors in Raid 0, and when I can clear enough diskspace I will put three WD 120's in Windows XP RAID 5.
I will lose 120 GB in capacity, but I will have 240 GB of storage that can survive the death of a hard drive.
Also Windows Raid can easily be migrated to new systems.
I will keep my 200 GB for emule Downloads, partion to image file backups. The 40 GB may go in my USB 2.0 enclosure.
I did run HDtach, but the problem is, it reads volumes, not partitions. Since software RAID works with partitions, HDtach cannot benchmark the array and will only bench the individual drives. Also, I am running in RAID 0 although I know it is risky. These drives are all brand-new so I'm willing to take the chance. Finally, I have seen CPU load as high as 15%, and I have yet to run any 3D games to see if the RAID array really decreases load times or increases FPS. I'm running a P4Northwood @ 3.25Ghz so I'm hoping that the CPU load will not be an issue. We'll see.
Ah I wasn't sure how low level HDTach went, apparently below the level of software raid.
I believe winbench has a transfer rate test?
Anyway a faster hard drive will improve load speeds, but won't affect FPS at all, since game's don't use the hard drive accept when loading. Higher CPU utilization would lower FPS, but again one its done loading the hard drive isn't utilized much.
The higher CPU might outway the faster transfers for loading speed, but I doubt it.
Don't you need an even number of partions for Raid 0?
We'll I've already published my Winbench results, but I think there is one other test I can try. It seems like the software RAID array is very flexible; we have seen it does not care about SATA/IDE, and it does not care about using just a portion of a drive. I do not think it wants an even number of drives or enclosures, either. As a point of interest, the SATA data cable fell out of my SATA drive earlier today,(while Windows was running!!) but it did NOT ruin the RAID array. Windows crashed, but rebooted just fine. Some of the data on the array was corrupted, but I bet that that would have been the case on a simple drive partition as well in that case.