I have a question I hope someone can help. This is specifically regarding breaking 2TB barrier in XP 32 bit. (I know I have the option of reinstalling the OS and redo my whole setup to 64 bit XP or Linux, but that's a whole separate discussion.)
With newer RAID cards, one can usually choose a subset of available RAID set space to create an OS volume.
Lets say I have total of 4TB of usable space. Let's assume all 4TB of usable space came from the same RAID 5 stack, so it's data-reliability is already accounted for.
I can do:
Volume A : 2TB
Volume B: 2TB
With 2 volume visible by XP.
My question is, what is wrong with then turning the two volumes into dynamic disks and then spanning them into one large 4TB virtual dynamic disk?
Every other post I've found online describes the disadvantages of dynamic disks, but I don't think it applies to this scenario because:
1. I am not planning to upgrade this OS for a while. Future VISTA-compatibility is NOT in the picture.
2. I am NOT changing my OS for this problem, so GPT, etc is out. I have to use MBR partitions.
3. I'm using the standard SPANNING dynamic disk, not the registry or hex hacked versions that could have issues with Windows Update reverting back.
4. Spanning the DD is not an issue, because behind the scene it's the same RAID 5 volume anyway.
5. DD allows me to easily expand the capacity in 2TB chunks if I need to in future. (i.e. do OCE 2 additional TBs, then create new 2TB volume, then have DD span to this new volume).
6. This seems to be the FASTEST solution (to bring online) yet! I don't have to do massive backup/restore (try doing typical backup/restore cycle when the array is 6TB and I'm trying to go to 8TB?!) I don't need to buy spare space to copy out then copy in -- besides, while copying-out, the "temp HDD" backup is likely non-RAID and is risky.
Now can someone tell me what's wrong with using Dynamic Disks this way? I can't think of the disadvantages, but I want to know of them before I do this.
PS: I have another PC that I setup to do spanning of Dynamic Disks. It's not RAIDed, and I simply spanned a *LOT* of old PATA harddisks and laptop disks together, so it's a 1.8GB volume I use mainly for backups. The DD solution works fine there, so I'm wondering what the problem with DD that everyone's screaming about?
I do believe this will work. If your RAID card can present multiple 2TB chunks to the OS (this is usually called LUN carving), then yes, you can create a spanned dynamic disk across all of them to get a >2TB volume.
My personal concern would be that this is nearly impossible to migrate out of, so when you upgraded to Vista/Server 2003/Server 2008 you would have to stay like this instead of going to a single basic disk on a GPT volume. However, you've stated that migration is not an issue, so go for it.
We have 1 customer using this exact configuration and I can tell you that is is VERY slow. Its on a 3ware controller too. My customer even has a different array to boot from. They have 10 1TB drives in a single drive letter in XP and its extremely slow. Its ok for them though because they dont need speed just size. One thing that helps (if you want to do it) is to figure out your total space in binary (not decimal) and divide that by how many carvings you will need when you set the carving size so they are all the same size.
That's surprising, because I would think the whole DD is resting on a RAID 5 stack, so the speed should be about equal to a RAID 5's speed.
However, I can see perhaps when you have a drive letter large enough, like 10TB in your customer's case, then NTFS itself starts to become bogged down; or maybe it is the dynamic drive spanning across 5x2TB makes the span-translation slow.
Anyhow, thanks for the feedback. I do care more about security than speed. I have a separate RAID-0 WD RAPTOR that I use for boot/programs. This is mostly for very large files that are typically accessed SEQUENTIALLY.