SERVER 2008 R2 install on RAID5 or RAID1

[Jaan]

Hi guys,

 

Im replacing some old servers (Yay). With the current setup we have a single RAID5 to which two virtual discs are setup. C: & D:

 

Im told that its better to have two RAIDS. RAID1 C: & RAID5 D: what do you guys have, and what are the benifits?

 

thx

[glennda]

Personally I setup OS on 2 x 72gb drives raid 1 and then data drives in raid 5 - seperates os and data

[soveryapt]

Evening

 

I have the 2 RAID setup as you have described. Basically, the benefits are that 2008R2 can be quite hungry on it's disk writes for the system, so to be safe that you're data access is minimally effected, you split it off onto another physical drive (well RAID).

 

I think the reason for the 2 different RAID standards is just simply to ensure you don't use up all the space on the server and it gives you a bit of additional room to play with (for instance, I have another RAID5 set with additional storage space I use for original photos etc (so when we take a load of photos at 18mp, I can size them sensibly for the general use and then store the full files (often RAW if I'm doing it) on the extra storage, again without effecting the general network storage).

 

But the basic thing is to keep the drives physically separated .. or at least that is my understanding of it and how I had it explained to me when I checked for confirmation with my supplier ..

[dhicks]

Im told that its better to have two RAIDS. RAID1 C: & RAID5 D:

 

A drive for the OS and a drive for data? Yes, although I'd go for RAID 6, which uses two disks for parity data rather than RAID 5's one. With today's larger, 2TB+ haddrives the chances of you having a read error somewhere on a disk are pretty good, and by the time you have 6 or 8 of them in an array you can bet you won't be able to read through all that data without a read error somewhere, so if a whole disk fail and you need to rebuild an array it's best to have two parity disks available. Alternativly, I understand Sun's ZFS has facilities for data redundancy, as does the succesor to NTFS due out with Windows 8 and as do various Linux file systems.

 

If you're buying a general-purpose server you'll want to run it as a virtual machine host, hosting several virtual machines. In that case, most of the main operating system probably gets loaded in to RAM on boot and never swaps out, the OS disk is probably just used to write logs. XenServer, for instance, insists on limiting itself to a 4GB OS partition, at least half of which is used by logs, so the main OS could probably be held in working memory or a RAM disk. If you have limited space in your server, or you just want to squeeze your monies-worth out of the harddrives you're putting in there, you could boot the OS off a USB stick or memory card and use all the RAID space for large harddrives.

[plexer]

Another +1 for RAID6

 

Ben

[sparkeh]

RAID 6 with a hot spare

[Jaan]

RAID 6 with a hot spare

 

i only have 6 bays!! LOL one server is 6x600gb sas and the other 6x300gb sas.

 

so would raid1 (2x600) and raid6 (4x600) without a hotspare be good? at the mo we have raid5 (6x174gb) inc a hotspare. the raid6 added parity would remove the need to have a hotspare would it not?

[Michael]

I've done both - RAID5, 2 partitions, but RAID1 OS and RAID5 data is my preferred choice.

 

It's all about failure rate. I'd rather have two separate RAID arrays rather than a single RAID array as re-installing the OS and restoring data (in a worse case scenario) it's going to create a lot of work.

[Arthur]

one server is 6x600GB SAS and the other 6x300GB SAS.

Enterprise-class SAS drives will generally be more reliable than standard SATA drives.

 

To give you an example, Seagate's Cheetah 15K.7 and Savvio 15K.3 SAS drives both have a non-recoverable read error rate of 10^16 bits, which means 1 sector in every 1.11022302 petabytes will be unreadable.

[MYK-IT]

Agree with RAID 1 for OS; as far as the other HDDs for DATA there are numerous options (and only you can decide what is best for you and its application).

 

Possible Configuration:

 

Server 1:

OS

RAID 1 (2x 600Gb) = 600GB

 

DATA

RAID 5 (No Spare) - 1.8TB (Only 1 HDD can fail and still continue) or

RAID 5 (+1 Spare) - 1.2TB (Only 1 HDD can fail, with 1 spare, and continue) or

RAID 6 (min 4 HDDs req) - 1.2TB (up to 2 HDDs can fail and still continue)

 

Server 2:

OS

RAID 1 (2x 300Gb) = 300GB

 

DATA

RAID 5 (No Spare) - 900GB (Only 1 HDD can fail and still continue) or

RAID 5 (+1 Spare) - 600GB (Only 1 HDD can fail, with 1 spare, and continue) or

RAID 6 (min 4 HDDs req) - 600GB (up to 2 HDDs can fail and still continue)

 

RAID 6 is slower (up to 50%) than RAID 5 for write operations due to the additional Parity strip but does have the unique redundancy level where by (any) 2 drives can fail and the data is still intact.

Even the likes of RAID 10 (or might be 0+1!) can cope with 2 (specific) HDDs failing but it cannot be any 2 HDDs.

 

It all depends on balance of performance, redundancy and overall storage space required.

For example using 2x 600GB RAID 1 for OS (600GB) may be considered overkill (and wasted resources).

 

Alternative configurations may be:

 

Server 1:

OS

RAID 1 (2x 300GB) = 300GB

 

DATA

RAID 5 (No Spare) - 1.8TB (Only 1 HDD can fail and still continue) or

RAID 5 (1 Spare) - 1.2TB (Only 1 HDD can fail, with 1 spare, and continue) or

RAID 6 (req min 4) - 1.2TB (up to 2 HDDs can fail and still continue)

 

Server 2:

OS

RAID 1 (2x 300GB) = 300GB

 

DATA

RAID 1 (2x 300GB) - 300GB (Fast Read/Write - Only 1 HDD can fail) and

RAID 1 (2x 600GB) - 600GB (Fast Read/Write - Only 1 HDD can fail)

[Jaan]

Agree with RAID 1 for OS

 

that post was epic!! thx for taking the time to respond like that! that helps me heap loads!

 

thx very much