Castlewood
asked on
Shall I use hardware RAID with Storage Spaces?
I'm trying to implement the "Storage Spaces" with Win2012 R2 with my Dell server, which comes with Hardware RAID for the DAS.
Question to you, what should I do with the hardware RAID? Shall I set up the hardware RAID 0 (or what?) and then set up the Storage Spaces in the OS?
Please help.
Question to you, what should I do with the hardware RAID? Shall I set up the hardware RAID 0 (or what?) and then set up the Storage Spaces in the OS?
Please help.
David
Set up RAID1. This protects against data loss in the event of a HDD failure or even an unreadable block on a HDD. (The "cost" for this is that your total usable capacity is 1/2 the total disk space). But in a perfect world this also doubles read speed since the controller will take advantage of two copies of the data and do read load balancing.
Using a hardware based RAID container with Storage Spaces defeats the purpose of using Storage Spaces. You can use the array but just use it as a set of JBOD. Storage Spaces is an advanced software RAID management system (in simple terms).
I would configure the HDDs/SSDs in the enclosure as Non-RAID disks and go thru the Storage Pool to setup redundancy and volumes.
For example, I have a DELL PE 730XD with a bunch of internal HDDs and several SSDs. It is setup as follows:
1. 2x disks for OS configured as a RAID 1 container, in hardware.
2. 8 disks, tagged as Non-RAID on the RAID controller which makes the disks visible to the OS
3. The 8 disks are configured in a Storage Services Storage Pool
4. The Storage Pool supports a volume that is usable by the OS for system services.
No hardware RAID on the 8 disks for data.
Storage Space overview: https://technet.microsoft.com/en-us/library/hh831739.aspx
Dan
I would configure the HDDs/SSDs in the enclosure as Non-RAID disks and go thru the Storage Pool to setup redundancy and volumes.
For example, I have a DELL PE 730XD with a bunch of internal HDDs and several SSDs. It is setup as follows:
1. 2x disks for OS configured as a RAID 1 container, in hardware.
2. 8 disks, tagged as Non-RAID on the RAID controller which makes the disks visible to the OS
3. The 8 disks are configured in a Storage Services Storage Pool
4. The Storage Pool supports a volume that is usable by the OS for system services.
No hardware RAID on the 8 disks for data.
Storage Space overview: https://technet.microsoft.com/en-us/library/hh831739.aspx
Dan
If you are using Storage Spaces for redundancy you may as well put the RAID controller into HBA/JBOD mode if it is available, would need to know what controller you have in order to tell how to do that.
A clarification ... use the hardware RAID for the Boot (C:) drive only.
Here is a nice how to .. note even IBM says to use hardware RAID1 for the O/S itself.
https://lenovopress.com/pdfs/1122/lp0064.pdf
Here is a nice how to .. note even IBM says to use hardware RAID1 for the O/S itself.
https://lenovopress.com/pdfs/1122/lp0064.pdf
ASKER
So I can simply disregard the hardware RAID (other than OS on C) as I will configure the HDD/SSD redundancy/volumes using Storage Spaces, correct?
Question to you, since no hardware RAID protection for the storage, how many failed hard drives does the Storage Spaces redundancy can sustain?
andyalder, We have Dell Perc 5e controller.
Question to you, since no hardware RAID protection for the storage, how many failed hard drives does the Storage Spaces redundancy can sustain?
andyalder, We have Dell Perc 5e controller.
The RAID level determines the amount of protection. (In perfect world) RAID1 protects against a single failure. RAID5 = 1 failure, RAID6= 2 failures. RAID10 = 1 failure per stripe.
But nothing is ever that simple. If you have RAID1/RAID5/RAID10 AND a bad block on the surviving disk, then you can still have partial data loss if a HDD fails.
I would just go with RAID6 for the bulk of the storage, and then a small RAID1 for scratch table space, index files, log files, and anything else write intensive This is a good bang for he buck since it gives twice the protection on the bulk of the data. Note RAID1 is also going to be much faster for reading than a single disk . All the more reason to use it for index files.
In event of a HDD failure on a RAID6 you can sleep at night because even though it might take days to rebuild if your system is busy, you still aren't all that exposed to catastrophic data loss in event a 2nd HDD disks. But any other RAID level (even RAID10) has risk of 100% data loss if a second drive fails. Again, RAID10 mirrors pairs of disks so for 100% data loss on a RAID10 then both drives in same pair have to die. Murphy's laws dictate that the 2nd disk that dies in a RAID10 will be the mirror of the first drive that failed ;)
The storage spaces filesystem adds a little more redundancy due to some checksums and such but in grand scheme of things it is effectively the same as described above. Now if you used ZFS, which is another filesystem then you would be better off, but that is not available for windows and is a Solaris/UNIX type thing.
But nothing is ever that simple. If you have RAID1/RAID5/RAID10 AND a bad block on the surviving disk, then you can still have partial data loss if a HDD fails.
I would just go with RAID6 for the bulk of the storage, and then a small RAID1 for scratch table space, index files, log files, and anything else write intensive This is a good bang for he buck since it gives twice the protection on the bulk of the data. Note RAID1 is also going to be much faster for reading than a single disk . All the more reason to use it for index files.
In event of a HDD failure on a RAID6 you can sleep at night because even though it might take days to rebuild if your system is busy, you still aren't all that exposed to catastrophic data loss in event a 2nd HDD disks. But any other RAID level (even RAID10) has risk of 100% data loss if a second drive fails. Again, RAID10 mirrors pairs of disks so for 100% data loss on a RAID10 then both drives in same pair have to die. Murphy's laws dictate that the 2nd disk that dies in a RAID10 will be the mirror of the first drive that failed ;)
The storage spaces filesystem adds a little more redundancy due to some checksums and such but in grand scheme of things it is effectively the same as described above. Now if you used ZFS, which is another filesystem then you would be better off, but that is not available for windows and is a Solaris/UNIX type thing.
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
Allow me to ask a side-tracked question...
Does the Storage Spaces provide de-duplication function?
Does the Storage Spaces provide de-duplication function?
ASKER
Never mind, Found de-dup was added on R2.