RAID1 vs backup software

[piratepast40]

I have 2-160G Seagate SATA drives. I've got 40G partitions on each with the idea that I can have my OS and program files on the first drives 40g partition and Flight Sim X on the beginning of the second drives smaller partition. I elected not to go with RAID 1 since I wanted fast access to the FSX files on the second drives smaller partition and at the beginning of the drive. Now I want to have decent backup capability for my OS, program files, and data files. Data files will be stored on the first drives larger (120g) partition.

I've got a funny feeling that I'm combining concepts without a complete understanding of them. I'd be happy to go with a RAID1 array for automatic backup but I really wanted that FSX program in the center (hub) of a drive for fast access. I'd be equally happy with some backup software that would give me a copy of my OS, program files, and data on the larger partition of my second drive. I'm kinda stuck on if I'm proceeding in the right direction or if I should change course.

This is a multi use machine. I use it for work as well as running flight sim. Can't afford to have the work stuff lost and don't want to go the pain of reconstruction in the event of loss of a drive (yes -I've been to that rodeo several times).

Any help is appreciated.

 

[piratepast40]

OK - I'm in a somewhat embarassing situation here - I think I may have asked a dumb question but I don't recognize it. I'm pretty thick skinned so if someone can tell me where to go find the answer, I'd be most appreciative.

Yes - I'm bumping my own thread!

 

[piratepast40]

OK - no response - that tells me that people are saying:
1) Is there a question here or what exactly is your question?
2)We can't tell what the heck you're talking about!
3)This has been answered so many times that I'm not going to waste my time until you do some research.

I've got 2-160g SATA drives and want to use them to replace the 80g PATA drive currently in my system. I want to set them up for speed, efficiency, and redundancy. To that end, I thought I should first set up a 40g partition on each and put my OS and program files on the first drives 40g partition. I would then set them up in a RAID1 configuration (the 40g partitions). The rest of the space would be used for data files that I could manually backup.

I've searched and searched but can't seem to find out how to RAID the partitions. Anything I google up points to Linux, not WinXP.

That makes my question: Can I even set up a raid array on virtual drives under WinXP or is it only for physical drives?

Note: I've allready tried to set up the array and it puts it on the larger partitions -I've lost the smaller 40g partitions. The full 160g shows up in bios (Promise fastbuild utility) as an array but in windows only as a 119g drive.

Thanks for any help

 

[Madwand]

RAID 1 is not an adequate substitute for a backup. If you have some malware/software/user error leading to data loss/corruption, then RAID 1 will quickly replicate that data to the other drive. A real backup should be external and disconnected. (Ideally multiple backups at different locations for really critical data.)

With the notable exception of Intel's Matrix RAID, RAID works at the drive level -- it takes multiple drives and transforms them into fewer drives. You can still split up your data on them into partitions though -- after the array has been created.

Another advantage of Intel's Matrix RAID is that it takes advantage of RAID 1 data duplication on reads (it cannot improve performance on writes). nVIDIA's (that I've seen) and some others don't do this.

I do not have an easy solution for you that will maintain both data integrity and performance, short of switching to an Intel solution. There are other solutions involving high-end RAID controllers, e.g. 3ware, which also "stripe" RAID 1 reads, but this generally requires a lot more money.

You should see the full 160 GB off the drive. That you aren't suggests that you might be running into a BIOS or OS service pack limitation -- look into updating these. Also watch out for the difference between GiB and GB -- Windows shows GB when it should show GiB. Look up GiB in Wikipedia.

Note also that Intel's Matrix RAID is not a magic solution for performance needs. If you partition the drive, and expect one part of it to be blazing fast and another part of it to be secure, consider that the time spent seeking back and forth between those parts can cost you significantly in performance.

For this, and other reasons, it's often recommended that the OS and data be on separate drives / arrays.

 

[piratepast40]

Thanks for the reply. I think I was just trying to do too many things.

Since my A8V mobo uses a VIA chipset and onboard Promise controller, the Intel solution isn't something I want to invest in.

With the notable exception of Intel's Matrix RAID, RAID works at the drive level -- it takes multiple drives and transforms them into fewer drives. You can still split up your data on them into partitions though -- after the array has been created

.

This was really the crux of my dilema. Since RAID works at the drive level and not on individual or specific partitions, I'll now need to go back and reformat the drives, set up the array, partition, and then reinstall the OS.

I understand that RAID1 will have a penalty in performance but thought I could overcome that by having the drive intensive application on it's own partition. I think I'll still do that but now understand that it won't be as advantageous as if I didn't array the drives. But just to make sure I understand, please let me know if I've got this right:

RAID works on the whole physical drive, not just individual partitions. This means that I can't (given my current hardware) have a non-raid partition. Is this correct?

Separate partitions for the OS and program files, and data files will still have an advantage in that the OS won't need to search the whole drive - just the smaller partition for information.
Is this correct?

Programs such as Ad-Aware, virus protection, and Norton system works will see only one drive (array) but will take longer to run since they now need to explore 2 identical data sets.
Is this correct?

Defrag of the OS/program file partition should be quicker since the program doesn't need to go over the whole drive area (unless I want it to).
Is this correct?
This question brings up another one concerning maintenance operations.
How do these programs address redundant data sets? Do they really search both, or just the main set and then write the data and make changes to the second set? It seems to me that the operation (defrag, cache cleanup, spyware and virus scans) would need to scan and perform operations on both data sets unless the second is completely overwritten. In either case, seems these operations would take double the amount of time. Any thoughts on that issue?

As far as backup. Once I've got things stabalized, I'll set file sharing back up with my laptop. Critical data will then reside on both. I also backup my current files about once a month to CD. Past years data is both archived and stored on CD's/data DVD's as well as residing on the desktop and laptop. The whole idea of the RAID1 array is so that I don't need to go through a total rebuild when a drive fails.

Sometimes I feel like my mind is wading through quicksand on this topic. I just can't seem to get it straight.

Thanks again for your help.

 

[SomeJoe7777]

What you're not understanding here is that RAID abstracts the concept of the physical drives. Once RAID comes into the picture, forget about physical drives. All that matters is how many logical drives have been created.

As an example, let's say you build a computer, and install one, single 160GB hard drive. You have one physical drive in the system. When the BIOS/driver goes and looks at the hardware, one logical drive is present. It's size is 160GB. You can partition the 160GB however you want, say into a 40GB and a 120GB partition, and install Windows on the 40GB partition and use the 120GB partition for data.

With RAID-1, it works like this: Two 160GB physical drives are installed in the system. In the RAID controller's BIOS, you tell the RAID controller you want to RAID-1 the drives. The RAID controller BIOS now takes the two physical drives (160GB each), creates a mirror, and presents one 160GB logical drive to the computer. Now, the system BIOS/driver and Windows all think that there's only one 160GB drive installed in the system. If you want to partition and format that as a 40GB and a 120GB partition, no problem. Everything after this point behaves exactly as the single physical drive situation described above. Nowhere in Windows or the machine's BIOS does it look like there are 2 physical drives. Everything looks just like there is only 1 drive in the system.

The difference is that if one of the physical drives dies, Windows believes that nothing has happened. It continues running, and is none the wiser. When you replace the failed drive with a new one, the RAID controller re-mirrors the other drive to the blank one and the array is redundant again. But you'll never see any of that happen in Windows.

Now, the other poster who noted "with the exception of Intel's Matrix RAID", is partly correct. RAID does work at the drive level. However, Intel's Matrix RAID also works at the drive level, just like every other RAID controller. The difference is that most consumer-level RAID controllers can only create one logical drive on the RAID array, and that logical drive has a fixed RAID level. (i.e. they can take 2 physical drives, make a RAID-1 logical drive, take 3 physical drives, make one RAID-5 logical drive, etc.) The Intel Matrix controller's secret sauce is that it can create multiple logical drives with a different RAID level on the same set of physical drives. For example, 2 physical drives @ 160GB each -> 1 logical RAID-1 drive (40GB), and 1 logical RAID-0 drive (240GB). Once that's done, the computer believes there are 2 drives installed, a 40GB and a 240GB. The 40GB is also protected from a drive failure (RAID-1). The 240GB is not (RAID-0).

RAID-1 typically does not have a performance penalty. It reads and writes just as fast as a single drive of the same size. Some RAID-1 controllers can do the reads from a RAID-1 somewhat faster than a single drive by interleaving the reads on the 2 physical drives.

So, to sum up, RAID controllers turn physical drives into logical drives, with the redundancy hidden behind the scenes. Windows and the rest of the computer only see the logical drives. There is no "searching of the redundant set of data" or anything like that. Windows and all your programs don't know anything about the redundant data.

 

[piratepast40]

Thanks so much. I was getting hung up on not seeing RAID and partition in the same discussion. This is my third attempt to set these drives up and wanted to make sure I fully understood what I was doing this time.

I was half expecting to see someone respond with "are you stupid?". Have to admit that I didn't think I was until I started this adventure !

It's funny how a slightly different explanation helps. I've been through the RAID FAQ sticky about 4 times now. The last link from storagereview, along with the responses here, finally got it hammered into my dense head!

As soon as the array is done formatting (again) I should be able to load XPpro on it and get back up and running.

Thanks again