Ghetto_Child
join:2004-05-30
Montreal, QC
2 edits |  [hard drive] RAID 0 Data Reliability vs Single Drive?
Ok I was thinking of setting up a 2 drive IDE RAID 0. My case only has space for 3 hard drives so before I try cooking up some design mod on my all metal 40lbs case I want to know how reliable a 2 drive RAID 0 IDE array is compared to a single IDE drive? I never have data corruption with normal use on my 1 drive system and I've only lost 1 hard drive in the past 7yrs and that was due to either the firmware or I accidently plugged a cable select jumpered drive to the slave end while my main hard drive was on the master end jumpered to master.
So 2 drive RAID 0 IDE data reliability and corruption vs a single drive IDE reliability. This is for personal use and I never save personal files on the boot drive. I always save custome work and settings if possible on a different partition at least. That way when the OS gets corrupted or needs to format I don't look any work and only the programs are erased and can be reinstalled fresh. |
|
ChiTang
Premium,MVM
join:2002-08-23
Alhambra, CA
| Reliablility if RAID 0 is minutely less than half of the reliability of a single drive. Put it in other way, RAID 0 is minutely more than double the probability of failure than a single drive.
--
I used to be indecisive, now I am not sure. |
|
Ghetto_Child
join:2004-05-30
Montreal, QC
2 edits | Is that just because there's 2 drives running as one (twice the probability)? Or does RAID 0 actually cause more corruption than a single IDE drive? |
|
Bigzizzzle
Premium
join:2005-01-27
Franklin, TN
| reply to Ghetto_Child
RAID 0 is bitchy and since it rights half the bits on 1 drive half on the other , data recovery is a major pain. I would put none essential stuff on the RAID 0 array ie windoes install, and programs all data on another drive either single or RAID 1 Mirroring. |
|
snipper_cr
join:2002-01-22
Wheaton, IL
clubs:
| I am doing what Bigzizzzle said.
I have 2 SATA drives in RAID 0 which is my primary boot drive. Its nice and fast and I have been doing RAID 0 (via Intel ICH5r onboard raid) for a couple years now with out a problem. And they are on MAxtor drives which historically die the fastest.
However.... I have 2 other harddrives configured in RAID 1 which I often back up my data too. In the event my main Raid 0 array fails... its all redundantly backed up.
What controller are you using to configure the RAID? Isnt there a RAID level which offeres some redundency with the speed of RAID 0? I think its called Raid 0+1 (but requires 4 drives)... might want to look into that.
--
Serenity Day - June 23rd 2006. You Can't Stop the Signal |
|
Ghetto_Child
join:2004-05-30
Montreal, QC
| yeah I know that RAID1 and RAID0+1 have fault tolerant redundancy. My PC case unfortunately only has room for 3 hard drives and 2 optical drives. I might be able to use 1 of the 2 3.5" floopy drive bays to hold a hard drive but it the biggest issue with that is the heat. It's a tightly concealed area so no air flow and at the moment I put a front access usb/1394/optical ports in there, will be annoying to lose that but it's an option.
So without changing anything I can only have 3 drives which give me RAID 0 plus 1 other drive. The controller I intend to use is a HighPoint RocketRaid133 PCI card v2.351 BIOS.
As for usage, I'm already safe on how I intend to use the drive, my policy with every hard drive is to make a minimum of 2 partitions and keep personal data and settings on 1 and programs on the other. That way formatting doesn't make me lose my personal files and settings. Second since it is going to be RAID 0 I already intended on keep important data off the entire RAID 0 array. |
|
koitsu
Premium
join:2002-07-16
Mountain View, CA
| said by Ghetto_Child  :... how reliable a 2 drive RAID 0 IDE array is compared to a single IDE drive? The risk is 50% greater (read: half as reliable). The more disks in a RAID 0 array, the higher the risk.
You need to remember that with RAID 0, your data is split across all drives. A retarded example would be, if you have a 4 disk RAID 0 array and write 4 bytes of data, each disk gets 1 byte.
Wikipedia has a fantastic article on all the RAIDs:
»en.wikipedia.org/wiki/Standard_R···s#RAID_0
In your case, since you're physically limited to 3 disks, my recommendation would be to use JBOD (keep all 3 disks separate from one another), unless you are *absolutely* in need of high read/write speed. In that case, keep 1 disk by itself (JBOD) and do RAID 0 across the remaining 2 -- but perform backups (to tape, DVD, whatever) on the RAID 0 array often (once a week at least).
said by Ghetto_Child :The controller I intend to use is a HighPoint RocketRaid133 PCI card v2.351 BIOS. I recommend you stay away from any/all Highpoint products. I'm not going to take the time to explain why because there's a good 8 years of history you can find on Google. The number of issues with HPT controllers under numerous operating systems (Windows, Linux, and the BSDs) is astounding. Data loss was quite common with their HPT360 and equally as bad with the HPT374. Some were tracked down to BIOS/firmware problems, others were engineering flaws which couldn't be worked around.
If money is a big concern and you need to purchase a RAID 0/1 controller, consider a Promise FastTrak controller. They only do SATA now, so if you need an older PATA controller, you can probably find one on eBay. Promise is inexpensive, and not the best, but their support is decent and the give out documentation on their controllers to developers (the idea being 1000 pairs of eyes is better than 1).
If you're planning on buying new disks for this project, I recommend you just buy SATA disks and get yourself a present-day Promise FastTrak SATA controller. PATA is on the way out.
--
Making life hard for others since 1977. |
|
Ghetto_Child
join:2004-05-30
Montreal, QC
2 edits | reply to Ghetto_Child
WHOA WHOA WHOA back up a bit. I have 3 hard disks only. I was going to add a RocketRaid 133 PCI card. I already have the card, just need warranty service on it because it got shorted by their's and Intel's and IBM's mistake. I connected the 4 LED pins on the PCI card to the 4 pins on the D850GB motherboard that were labeled SCSI LED and when I did that the system couldn't boot, so I tried the pins in reverse order and still couldn't boot. Then I removed the pins and booted but discovered the card was now unstable.
Anyway I never ran the card in an array yet and had it a few months. I updated the RocketRaid card's BIOS to V2.351 as soon as I got it and that had no problems. The chipset on it is HPT372A it's a 4 drive 2 IDE port controller. I don't think I can do a JBOD without hindering speed because I've never heard of RAID0 on the controller and the backup on a seperate 3rd disk connected to the motherboard controller?
I don't need extra storage space I was looking to maximize speed. 2 drives RAID0 running the OS on the primary partition which will be on the outer most 10GB rings of each hard disk I figure that would be amazingly fast for IDE. Afterall hard disks spin at constant RPMs and the circumference of a circle is 2(Pi)R. Since 2(Pi) (pronounced two pie) is constant through the whole disk then your actual speed is a multiple of R the radius from disc center to your position on the disk. Same applies to any spinning medium at constant RPM.
A hard drive is about 4" wide, the casing takes off 1/8" or 0.125" on each side. The spindle in the middle is about 1.25" so that leaves us with 2.5" of writing surface diameter on the platter or 1.25" writting surface radius. That means at your outer most ring of a desktop hard drive your surface velocity is 1.25 times whatever you have at the first inner most ring. I think this surface velocity is the angular velocity.
However I think I recall users getting a much greater speed boost than 1.25 times if my memory doesn't fail me so I'm not too sure how this fits in now, but definately your drive mechanics are performing 1.25 times more efficiently. |
|
koitsu
Premium
join:2002-07-16
Mountain View, CA
| said by Ghetto_Child :The chipset on it is HPT372A it's a 4 drive 2 IDE port controller. I don't think I can do a JBOD without hindering speed because I've never heard of RAID0 on the controller and the backup on a seperate 3rd disk connected to the motherboard controller? You can run any combination of anything you want. Example:
Config #1
* Two (2) drives in RAID0 on RAID controller
* One (1) drive on mainboard controller
Config #2
* Two (2) drives in RAID1 on RAID controller
* One (1) drive on mainboard controller
Config #3
* Three (3) drives as JBOD on RAID controller
And so on. This won't affect performance or impact your system at all, unless there's busmastering problems or IRQ sharing problems (depending on how old the board is, it may not matter, since APICs are common now).
said by Ghetto_Child :I don't need extra storage space I was looking to maximize speed. 2 drives RAID0 running the OS on the primary partition which will be on the outer most 10GB rings of each hard disk I figure that would be amazingly fast for IDE. {snipp} I'm incredibly confused by the analysis you're doing here. "Outer-most rings?" When you create a RAID array, you do not tell the array what physical portion of the disk you want to use; the controller will use the entire disk, not a portion of it.
Hard disks are like CDROMs in the sense of how the data is written -- it is written from the inside (e.g. closest to the spindle) to the outside (e.g. the edge of the platter). This is not something anyone bothers to take into mind when choosing a RAID type or when making an array. 
RPMs, naturally, plays a large role in regards to how fast you can syphon data to/from the disk across the bus. You'll get better overall speed with a 15krpm disk than you would a 10krpm vs. 7200rpm vs. 5400rpm. On-disk cache also plays quite a large role, as does stripe size (in the case of RAID 0). These are the major factors in making a RAID 0 array.
In summary, you *will* achieve very high read/write performance with a RAID 0 array, simply because the data is evenly split across all the drives in the array (in your case, two drives). But as mentioned, you increase risk of failure using RAID 0 by 50%.
My point about RAID 0, though, is that as long as whatever data is on the RAID 0 array can be lost permanently (1 drive failure is all it takes), or can be backed up somewhere regularly/often, then using RAID 0 is perfect for what you want.
--
Making life hard for others since 1977. |
|
Keizer
I'M Your Huckleberry
Premium,MVM
join:2003-01-20
| reply to Bigzizzzle
said by Bigzizzzle :RAID 0 is bitchy and since it rights half the bits on 1 drive half on the other , data recovery is a major pain. A major pain in what way??? I have been running a RAID 0 array for over two years. I had one issue in that time frame. I lost one of my drives. I simply replaced the drive with a spare that I had which was the same brand, and size. I then went into the RAID setup, deleted my array, recreated it with the same name, and installed my most current image file using True image. I was back up and running in 15 minutes tops!!
I only agree with your statement if you don't back up. Imaging is a must if you run with a RAID 0 array. |
|
Ghetto_Child
join:2004-05-30
Montreal, QC
| reply to koitsu
said by koitsu :I'm incredibly confused by the analysis you're doing here. "Outer-most rings?" When you create a RAID array, you do not tell the array what physical portion of the disk you want to use; the controller will use the entire disk, not a portion of it. Not entirely true, a hard disk uses the inner most portion of the PARTITION not the of the disk. You see if you split your drive into more and more partitions then you'll have smaller and smaller pieces of logical drives. Also if you restrict a partition to the outer most edges of the disk then you will be restricting the partion to use the fastest rings of the disk. I've done it before as a test, I used a Computer Management to partition the disk, I made the extended partition first as big as I wanted to reserve it and then I made the primary partition the rest of the way. The result is the primary partition was at the end of the disk and the secondary partition at the beginning. Any retail partition manager can do this even easier.
I was then able to either delete the extended partition or remake it with logical drives whatever sizes I want. This resulted in the primary partition being the fastest partition and windows installation from as far as Win95 all the way up to WinXP doesn't have a problem formatting and installing on a pre-made partition regardless of it's location on the disk.
Now take this idea and apply it to a 300-500GB drive. If you make the extended partition first as large as possible and reserve just 10GB-40GB unpartition space. Then make the primary partition in the rest of the space, windows computer management will place the primary partition in whatever block of space is left after the secondary partition, literally and physically.
said by koitsu :Hard disks are like CDROMs in the sense of how the data is written -- it is written from the inside (e.g. closest to the spindle) to the outside (e.g. the edge of the platter). This is not something anyone bothers to take into mind when choosing a RAID type or when making an array. Yes hard disks write from inside to out, but if you restrict a partition's size and location to being the outer most rings then you will be reading and writting from the inner most ring to the outer most ring OF THAT OUTER MOST PARTITION and not the whole disk. I've tested this several times in all sorts of ways using computer management and it works perfect. I don't know if FDISK in dos will let you do that but you will need to make a non-dos partition to reserve the space instead of making an extended partition first. FDISK doesn't allow you to make an extended partition unless a primary partition exists first.
Here's an example, you have a 100GB unpartitioned and unallocated drive. You tell windows to first make an extended partition, you tell it to use 80-90GB for the extended partition. Drive whirs for a bit and does it's allocating. Now you have 10GB-20GB of unallocated space left at the end of the drive. You tell windows to make a primary partition using that. Voila, primary partition in the last 10GB-20GB of your physical drive ensuring that you now have the fastest rings of the disk allocated to the primary partition. |
|
Ghetto_Child
join:2004-05-30
Montreal, QC
1 edit | reply to Ghetto_Child
Well I don't have faulty or overclocked parts in my system. I do get system freezes forcing me to switch the system power off instead of a shutdown every couple of months but that only occurs when testing out software bugs and when I havn't turned off my PC in 2-3months. Power failures here happen at once per season. Those are the only things I will have affecting my data. In 3 years all the crashes, power failures and system freezes I've had never gave my hard disk a bad sector, never caused me any data loss and all file errors were fixed upon reboot. Usually the problem was incorrect reported sizes, or some other minor discrepancy like that.
Would these types of errors cause the array to fail? Would these types of crashes cause data loss on RAID0? Or is the data loss for these specific situations the same as a single IDE drive? |
|
Bigzizzzle
Premium
join:2005-01-27
Franklin, TN
| reply to Keizer
cause raid generates a lot of faulty errors and condition of the drive readings. Plus with the average drive the Firmware on the hardrive it self sometimes causes conflicts. Take for example recently WD on its newer drives had to update its firmware like many manufacturers that would exhibit false messages ie RAID drive degradation. When in fact the drive was fine. I dislike RAIDs in non enterprise environments; its another example of "you pay for what you get." |
|
