HOWTO: Software RAID 1/5 under XP Home/Pro

raid 5 on 2 disks is beyond pointless. to do HW raid 0+1 or raid 10 you'll need 4 disks minimum.

can you post some HD tach graphs when you get a chance.

That's a pretty damn cool hack! Thanks for the link!

I generally build file servers at work like this...

disk 1 + disk 2 = Raid 1 (C drive, boot win2k3 drive)
disk 3 + disk 4 + disk 5 = Raid 5 (D drive, data storage drive)
disk 6 + disk 7 = hot standby

nice simple fast redundant configuration... generally on IBM xServers.

SoNgMaN wrote:raid 5 on 2 disks is beyond pointless. to do HW raid 0+1 or raid 10 you'll need 4 disks minimum.

can you post some HD tach graphs when you get a chance.

No, you don't do RAID 5 w/ 2 disc. (Althought it is possible under Linux, you can even do RAID 5 w/ just ONE HD!!! )

And no, this doesn't do RAID 10 or 1+0 either.... (but once again, this can be done in Linux w/ even a single drive)

HD tach graph doesn't draw the graph by partition but by HD only. I am still looking for some "light/freeware"* tool to plot some nice graph for my D: & E:. I am currently just using some simple tool (I think is call Disk speed or drive speed) that write a bunch of files on the partition and measure the speed that way....

* Don't feel like to uninstall the apps after testing and leave behind some crap.

sleepyguy wrote:I generally build file servers at work like this...

disk 1 + disk 2 = Raid 1 (C drive, boot win2k3 drive)
disk 3 + disk 4 + disk 5 = Raid 5 (D drive, data storage drive)
disk 6 + disk 7 = hot standby

nice simple fast redundant configuration... generally on IBM xServers.

Do you work for IBM? This config just looks familiar to me.. hehe

Pretty neat setup. I can see the "Pro" part of it. I am thinking about the "Con" side ...

1. This should work well if only the RAID-0 *or* RAID-0 partition is being accessed. The performance should drop (in half ?) if they are both being used at the same time.

2. Can Norton Ghost (DOS mode) 'see' the arrays ?

willy wrote:Pretty neat setup. I can see the "Pro" part of it. I am thinking about the "Con" side ...

1. This should work well if only the RAID-0 *or* RAID-0 partition is being accessed. The performance should drop (in half ?) if they are both being used at the same time.

2. Can Norton Ghost (DOS mode) 'see' the arrays ?

Well, if there are frequent access to both RAID 1 & 0 partitions, yes, there will be a performance hit, but then it is same as have one HD and partition for D & E anyways... In my case, since it is MCE, most of the time I will not have access to both D & E at the same time (unless I am running slide show meanwhile the schedule recording kicks in. But then loading up a single pic vs streaming video to the HD, that's nothing...)

About Ghost in DOS mode, haha, good luck. Does Ghost work for software RAID array? anyone?

I *do* know if this is NOT windowz, but Linux, I can easily resize the RAID array, if I have defined the LV for RAID in my VG. I can add/remove PE to my LV and use lvextend/ext2online/mdadm to resize it

Finally I have some time again to test out another config...

Both 250GB on the same IDE channel vs each 250GB on individual IDE channel

The results? Very interesting, and it wasn't what I have expected.

D: is my Mirror partition
E: is my striping partition
My OS drive is on a seperate SCSI channel.

Can you guess which test is which config?

Test #1
[QUOTE]Starting Batch Create File Bench...

48 MB; D:\test; 50331648 bytes; 1328 ms; 36.145 MB/s
52 MB; D:\test; 54525952 bytes; 1000 ms; 52.000 MB/s
56 MB; D:\test; 58720256 bytes; 1141 ms; 49.080 MB/s
60 MB; D:\test; 62914560 bytes; 1219 ms; 49.221 MB/s
64 MB; D:\test; 67108864 bytes; 1312 ms; 48.780 MB/s
68 MB; D:\test; 71303168 bytes; 1422 ms; 47.820 MB/s
72 MB; D:\test; 75497472 bytes; 1531 ms; 47.028 MB/s
76 MB; D:\test; 79691776 bytes; 1656 ms; 45.894 MB/s
80 MB; D:\test; 83886080 bytes; 1766 ms; 45.300 MB/s
84 MB; D:\test; 88080384 bytes; 1859 ms; 45.186 MB/s

Create Batch File Bench ended

Starting Batch Create File Bench...

48 MB; E:\test; 50331648 bytes; 641 ms; 74.883 MB/s
52 MB; E:\test; 54525952 bytes; 609 ms; 85.386 MB/s
56 MB; E:\test; 58720256 bytes; 672 ms; 83.333 MB/s
60 MB; E:\test; 62914560 bytes; 734 ms; 81.744 MB/s
64 MB; E:\test; 67108864 bytes; 797 ms; 80.301 MB/s
68 MB; E:\test; 71303168 bytes; 844 ms; 80.569 MB/s
72 MB; E:\test; 75497472 bytes; 922 ms; 78.091 MB/s
76 MB; E:\test; 79691776 bytes; 969 ms; 78.431 MB/s
80 MB; E:\test; 83886080 bytes; 1047 ms; 76.409 MB/s
84 MB; E:\test; 88080384 bytes; 1078 ms; 77.922 MB/s

Create Batch File Bench ended[/QUOTE]

Test #2
[QUOTE]Starting Batch Create File Bench...

48 MB; d:\test; 50331648 bytes; 1219 ms; 39.377 MB/s
52 MB; d:\test; 54525952 bytes; 1281 ms; 40.593 MB/s
56 MB; d:\test; 58720256 bytes; 1422 ms; 39.381 MB/s
60 MB; d:\test; 62914560 bytes; 1562 ms; 38.412 MB/s
64 MB; d:\test; 67108864 bytes; 1719 ms; 37.231 MB/s
68 MB; d:\test; 71303168 bytes; 1953 ms; 34.818 MB/s
72 MB; d:\test; 75497472 bytes; 2016 ms; 35.714 MB/s
76 MB; d:\test; 79691776 bytes; 2141 ms; 35.497 MB/s
80 MB; d:\test; 83886080 bytes; 2297 ms; 34.828 MB/s
84 MB; d:\test; 88080384 bytes; 2438 ms; 34.454 MB/s

Create Batch File Bench ended

Starting Batch Create File Bench...

48 MB; e:\test; 50331648 bytes; 609 ms; 78.818 MB/s
52 MB; e:\test; 54525952 bytes; 719 ms; 72.323 MB/s
56 MB; e:\test; 58720256 bytes; 766 ms; 73.107 MB/s
60 MB; e:\test; 62914560 bytes; 797 ms; 75.282 MB/s
64 MB; e:\test; 67108864 bytes; 844 ms; 75.829 MB/s
68 MB; e:\test; 71303168 bytes; 828 ms; 82.126 MB/s
72 MB; e:\test; 75497472 bytes; 1000 ms; 72.000 MB/s
76 MB; e:\test; 79691776 bytes; 1141 ms; 66.608 MB/s
80 MB; e:\test; 83886080 bytes; 1234 ms; 64.830 MB/s
84 MB; e:\test; 88080384 bytes; 1141 ms; 73.620 MB/s

Create Batch File Bench ended[/QUOTE]

Well.. Test #1 is actually BOTH drives on the SAME IDE channel, and appear to have better performance than having the drives runs in seperate channels. Very interesting eh? This is done on the same machine, and I only swapped cable between tests. One would expect each drive on seperate channels will run faster, but seems like this shown otherwise.

My initial guess is somehow when the OS needed to communicate w/ each IDE channel, it takes some time to set up the communications per channel.

Anyways, I have found my answer.. now is time for me to continue my work...

So your tests say, putting both HDs on a single channel is more advantageous than having each on their own channel? Is this valid for RAID setups only?

Also, I know this question has been asked to death, but i'm going to ask again to raid or not to raid. For everyday how users, since sustained disk activity is rare, most users will not see the benefits of striping.... but 1 second faster is still faster right? How applicable is that in everyday use? Do you "feel" the difference?

Not surprising. And nothing to do with taking time to setup communications per channel.

The reason is because IDE devices do not have its own direct communication to the controller. Each IDE channel, needs to communicate to the controller via the IDE bus line. So you wouldn't get faster performance by putting two harddrives on seperate channels. Because each channel needs to share the same IDE bus line to communicate with the controller, using more channels only serve to create a bottleneck in peak data traffic. However, using one channel means the bus only needs to deal with one channel and therefore the data throughput would be faster. This is a drawback to the parallel bus architecture. For this reason is why even SCSIs can have the same drawback.

Nothing to worry about though. This is already a well known fact for manufacturers which is why the path to data throughput is evolving towards serial technology. Hence the already in existence Serial ATA drives and the forthcoming SAS (Serial Attached SCSI). This will allow each drive to communicate directly to the controller via its own fast point to point connection. This is where we will see the performance gains.

IceMan77: I feel the difference! I'll never go back to single...

IceMan77 wrote:For everyday how users, since sustained disk activity is rare, most users will not see the benefits of striping.... but 1 second faster is still faster right? How applicable is that in everyday use? Do you "feel" the difference?

Instead of adding a single large HD, I am going to add a 2xHD RAID-0 array to my DV machine to handle all my DV-to-DVD conversion. I 'believe' (I don't have hard data to prove) RAID-0 should help dealing with those multi-GB files somewhat.

Cafe_333 wrote:
IceMan77: I feel the difference! I'll never go back to single...

Doh.... That's all i needed to hear, now i gotta see for myself.... Now before I tear apart my setup and RAID 0 my box.

What is your software/hardware setup like? How do you use your RAIDed box? I don't do video editing. Mostly using many apps, and some gaming. I wonder how much I will "feel" the difference, cuz it's all about the feeling hehe

I also use DriveImage to ghost my image from time-to-time (great software by the way). After I RAID, will i still be able to restore? DriveImage is Caldera Dos based.

IceMan77 wrote:So your tests say, putting both HDs on a single channel is more advantageous than having each on their own channel? Is this valid for RAID setups only?

Also, I know this question has been asked to death, but i'm going to ask again to raid or not to raid. For everyday how users, since sustained disk activity is rare, most users will not see the benefits of striping.... but 1 second faster is still faster right? How applicable is that in everyday use? Do you "feel" the difference?

Well, I think the test shown advantageous if a OS needed to write to both drive alot (or the same time) you will have (but may not actually 'seen') some improvements.

the question to RAID or not to RAID is really should be "to what RAID or not to RAID" Because each RAID levels are target for different situtations. You could do RAID1 (mirror) which has nothing to do w/ performance but to protect your DATA. And RAID0 could be use if you needed a large storage space across 2 HDs (e.g. you have 2 smaller HDs)

So my answer to you is depends on how much you value your data.

The Multi-RAID types on the same drives you're referring to is Intel RAID Matrix.

Kind of neat in theory, but nothing I don't think i'd want to implement.

'[H wrote:ackerK']Well, I think the test shown advantageous if a OS needed to write to both drive alot (or the same time) you will have (but may not actually 'seen') some improvements.

the question to RAID or not to RAID is really should be "to what RAID or not to RAID" Because each RAID levels are target for different situtations. You could do RAID1 (mirror) which has nothing to do w/ performance but to protect your DATA. And RAID0 could be use if you needed a large storage space across 2 HDs (e.g. you have 2 smaller HDs)

So my answer to you is depends on how much you value your data.

I'm RAIDing only for performance reasons. All my data files reside on my file server. Like I said my main system is used mainly for opening and closing various apps and some gaming.

I'm just like yourself, mostly apps and some gaming! I hardware raid-0 my two harddrives so even the OS is raided, and offload anything I want to keep to a file server. You *will* feel the difference for sure!

Everything opens/reads faster, everything writes faster, and everything loads faster. When I click to open an app, it's at a blink of an eye. Made me wonder why I didn't do it sooner.

If you want to see some benchmarks, check out the following link. Too lazy to look for direct comparisons as this review was for something else, but they did compare a Single Drive versus Raided drives. Read and Write speeds are literally doubled. For raid 0 versus a single drive, just look at the 'blue' and 'cyan' colors.
http://www.hardcoreware.net/reviews/review-283-2.htm

IceMan77 wrote:Doh.... That's all i needed to hear, now i gotta see for myself.... Now before I tear apart my setup and RAID 0 my box.

What is your software/hardware setup like? How do you use your RAIDed box? I don't do video editing. Mostly using many apps, and some gaming. I wonder how much I will "feel" the difference, cuz it's all about the feeling hehe

I also use DriveImage to ghost my image from time-to-time (great software by the way). After I RAID, will i still be able to restore? DriveImage is Caldera Dos based.

Tom's Hardware had an article on doing software RAID-5 with WinXP a few months ago... if interested, you can look that up.

Cafe_333 wrote:I'm just like yourself, mostly apps and some gaming! I hardware raid-0 my two harddrives so even the OS is raided, and offload anything I want to keep to a file server. You *will* feel the difference for sure!

Everything opens/reads faster, everything writes faster, and everything loads faster. When I click to open an app, it's at a blink of an eye. Made me wonder why I didn't do it sooner.

http://www.hardcoreware.net/reviews/review-283-2.htm

Lol, I'm getting all excited just reading about it!!!.... it's too bad, I just don't have the time to rebuild my box till next month... arghhh.

I'm guessing you're using a dedicated card, since Win2K/XP can't raid 0 their own partition....

Now if only someone can verify ghost or drive image can restore to a RAID 0 drive, there's no excuses for implementing RAID!

Don't mean to hijack the thread... but anyway, if your mobo has Raid, you can do it that way too. I've worked with onboard & dedicated and both work just as well. The latter has less overhead but you won't really see the real benefit of this unless you were doing high burst transfers. Otherwise for just apps and games, you'll be pretty happy with the performance of either. As for ghosting, I have not tried it myself with striped disks nor do I know if it's possible, but a quick lookup on the Symantec website says that Norton Ghost 9.0 supports it. Beyond that I couldn't tell you any more. Enjoy!
http://service1.symantec.com/SUPPORT/po ... =&osv_lvl=

IceMan77 wrote:Lol, I'm getting all excited just reading about it!!!.... it's too bad, I just don't have the time to rebuild my box till next month... arghhh.

I'm guessing you're using a dedicated card, since Win2K/XP can't raid 0 their own partition....

Now if only someone can verify ghost or drive image can restore to a RAID 0 drive, there's no excuses for implementing RAID!