• Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 489
  • Last Modified:

Which type of Raid to use???

I have 2 HP DL 380 servers both with 3 x 70 gig drives and 3 empty slots, I can fill these if required.
My question is what Raid Config do you recommend for the following:

1 server as a 2003 R2 Server
Domain Controller
Used purely as a DC in an existing domain, it will NOT be used for file shares or anything else it will just sit there and function purely as a DC & GC with DNS & DHCP.

1 server as a 2003 R2 Server
This will NOT be a DC,
It will however be a 2003 Exchange server,
Again no file shares or anything else, it will just sit there and function purely as a 2003 Exchange server.

These servers are new and have as yet not being built, Im aware of having to run forest prep on the existing schema master for R2.
I would however love to hear your recommendations on the most appropriate raid configs for both servers.
0
rpartington
Asked:
rpartington
  • 5
  • 4
  • 2
  • +6
2 Solutions
 
rindiCommented:
Use raid 1 with an online spare. This should be the easiest to setup, you get good failover, the speed is good, and since they are only AD services servers with no other functions, the usable space of 70 GB is more than adequate.
0
 
arnoldCommented:
I agee with Rindi's recomendation of raid 1 for the OS with a hot  spare in the AD server.

As to the exchange server provided it does not have to run on a DC, two drives in RAID 1 for the OS and applications, the remaining four in a striped mirror configuration Raid 1+0 for the exchange's datafiles.  while there is no hot spare, you can keep one or two  drives in a cabinet just in case one fails.
Using a RAID 5 config (3 drives) for the exchange data would have been my suggestion a few years back, the performance hit on the exchange server should a drive fail and the time it takes to rebuild have since changed my mind. i.e. the failure in a drive will likely occur near the manufacturer stated usage data/spec.  The three drives will likely be of the same "age" when one fails.  You would then have to hope that the rebuild completes before any other drive fails/encounters errors.
0
 
scrathcyboyCommented:
follow the above RAID 1 specs.  It is the MOST reliable.  If you use RAID 5 or 0 you are asking for failures down the road.
0
Technology Partners: We Want Your Opinion!

We value your feedback.

Take our survey and automatically be enter to win anyone of the following:
Yeti Cooler, Amazon eGift Card, and Movie eGift Card!

 
prashsaxCommented:
Just a thought for exchange.

If you have many users, then might feel the speed performance given by RAID 1.

If you plan to backup your exchange Database regularly, go for RAID 5.

It has good speed, and fairly good protection.
0
 
rpartingtonAuthor Commented:
only 150 users
0
 
Darwinian999Commented:
Take one disk out of the DC and put it in the Exchange Server.
Configure the DC servers 2 disks as RAID 1. This will be ideal for a DC with 150 users.

Configure the Exchange servers 4 disks as RAID 10. This is ideal for an Exchange server - compared with RAID 1, you'll have twice the read performance, twice the write performance, no loss of performance if 1 disk fails, and the ability to survive up to two simultaneous disk failures (if they're the right two). If you need more space or more performance, add another 2 disks into the RAID 10 set (giving you 3 times the performance of RAID 1, and the ability to survive up to 3 simultaneous disk failures), or else use the additional 2 disks as a RAID 1 set and put the Exchange logs  & queues on it.

Don't even think about RAID 5 - it has poor write performance which only get worse when a disk fails - not something that you want for a DC or an Exchange Server.
0
 
David_FongCommented:
A 4 * 72GB disk Exchange server gets 10 GB RAID 10 for the OS/pagefile, 10 GB RAID 10 for the transaction logs and 180 GB RAID 5 for the data with maybe carving the RAID 5 bit into three 60GB partitions for public, private and recovery/repair DBs if I set it up.

If a disk is down then you ask the users to think twice before editing a fle. The poor write performance of RAID 5 can be ignored since write is performed on the 1+0 bit of the disks with an Exchange lazy process transferring it to the RAID 5 data volume. And you buy write cache to offset the problem anyway.
0
 
IPKON_NetworksCommented:
I totally agree with David Fong.

RAID 1 for DC server
RAID 1 for OS and RAID 5 for Data  with a floating Hot Spare disk between the two volumes. Spend your extra money on battery backed cache for the controller on the Exchange server and you have the best of all worlds. Has been the standard config for data servers for many many years as it is the best solution for the money.

Barny
0
 
Darwinian999Commented:
Something that's been done for "many many years" isn't always the best thing for the here-and-now. RAID 5 was an acceptable compromise when the cost of storage was very high. Nowadays, disks are cheap, so RAID 10 is a far better choice. For the cost of one extra disk (3 disk RAID 5 v's 4 disk RAID 10) you get the same usable capacity with twice the write performance, the ability to survive two simultaneous disk failures, and no loss of read or write performance when a disk fails.
0
 
David_FongCommented:
I do prefer RAID 10 but with only 72GB disks to play with they are likely to run out of space and as the controller allows mixing of RAID levels on the same disks why not use that functionality. There's also the option of the two RAID 10 sets for OS and transaction logs and a third RAID 10 set for the data, then if they run short of space they can migrate the RAID 10 to RAID 5 on the fly without converting the OS and logs virtual disks.

Quite frankly it would make more sense to take all three disks out of the DC and put them in the Exchange server along with the RAM and PSU, it wouldn't have much work to do since GC queries would be resolved locally instead of accross the LAN if you dcpromo it and make it GC server. DNS and DHCP aren't enough load on the server to worry about.
0
 
Darwinian999Commented:
Although I run my own Exchange Server as a DC (home/test network that only supports a few users), it's not something I would recommend (nor do Microsoft). The optimisations, recovery strategies, etc for DC's and Exchange servers are quite different.

If you wanted to go down the path of hardware consolidation (not a bad way to go), then a good option would be to run 3 virtual servers on the consolidated hardware - 2 DC's with minimal resources, and an Exchange server with the majority of the resources. Microsoft Virtual Server and VMware Server are both free. I'd still configure the physical disks on the consolidated server hardware as a single RAID 10 set though, to get the performance, redundancy and availability... :)

I'm going to be doing this myself in a few weeks, once I get some new hardware to run a bunch of VM's on. It will be running RAID 10 (as it does now) and will have VM's for 2 DC's, 1 Exchange server and possibly a linux server or two (for firewalls, traffic monitoring, web servers etc).
0
 
rpartingtonAuthor Commented:
on a raid 1+0 raid 10 for the exchange server what is the recommended Stripe size for the logical drive 128kb???
0
 
arnoldCommented:
128KB??

Could you recheck the question above and clarify?  What stripe?

For which logical drive are you asking about the space allocation?  
0
 
Darwinian999Commented:
I ran some brief unscientific performance tests on some DL385's, some with 6 x U320 disks in a RAID 10 set, and some with 4 x U320 disks in a RAID 10 set, with 64K and 128K stripe sizes. 64K was a bit faster. I plan to do some proper tests in a few weeks comparing IO performance of various numbers of SAS disks in RAID 1, 5 and 10 with various stripe sizes connected to a DL385 with a P600 RAID controller (256MB cache).
0
 
rpartingtonAuthor Commented:
Thats interesting Darwinian
as Im just configuring 2 HP DL380s
and their Array Config utility defaults to 128KB for stripe size when you select Raid 1+0, obviously you have the option of selecting various others but this is when selecting
4 x 72gig drives for the exchange in a RAID 1+0.

Arnold I have 2 servers with a grand total of 6 HDs.
To get to this stage on the exchange I took 1 drive from the DC and popped it into the Exchange so I now have 4 x 72 gigs in the Exchange.

I can if need be purchase more drives but dont really want to if I can help it.

Neither have being configed completely as Im waiting for the courier to deliver the Server 2003 R2 Cd and an email supplying the Key Codes etc so at the moment both servers are just sat displaying the HP Array Config utility.
0
 
pyroman1Commented:
@Darwinian999-
You mentioned this:
"...you get the same usable capacity with twice the write performance, the ability to survive two simultaneous disk failures..."
and it made me wonder, when did you last check this?  I currently run a RAID5 at home with 4 SATA2 Seagate drives and my burst speed is over the rated 300MB/s with avg speed around 117MB/s.  Are you suggesting I could get 200+MB/s avg speed with RAID10?  I used HDTach to measure performance.  One other note, since you did bring up the possiblility of two drives failing, the mirror image could be the other failed drive which means you lost all of your data.

My suggestion is to go with RAID5.  In my experience it is faster and provides more storage space.
0
 
Darwinian999Commented:
pyroman1: If you find RAID 5 faster, good luck to you. Especially when a disk dies. There's a big difference between the I/O patterns of a single-user home system and a multi-user server.

Each to their own. There's a strong camp of RAID 5 followers that will probably stick with RAID 5 until they or their RAID 5 sets die. As for me, I move with the times - disks are relatively heap nowadays and there's too many advantages to RAID 10 to ignore it... :)
0
 
pyroman1Commented:
I use RAID5 at work as well, I just can't comment on the speed there as I really saw no need to check it.  Perhaps now I will.  My point was simply to inquire as to when you determined it was not so fast.

The chance of multi-disk failure is extremely low, therefore the "benefit" of having a system that can recover from one is not that great.  In addition, as I mentioned previously, the chance of losing the second drive which has a copy of the first that failed with only 4 disks is very close to that of losing two disks period thus causing complete loss of data, just like in RAID5.

Here is more information about RAID which discusses RAID5 and RAID10.
http://en.wikipedia.org/wiki/RAID

Bottom line, everyone is going to have a different opinion.
0
 
rpartingtonAuthor Commented:
Many thanks for your opinions and recomendations, at the end of the day as pyroman1 correctly stated its all down to personnel preferance.
I just felt that Arnold & Darwinian999 recommendations suited my situation slightly better.
Again many thanks for all your help.

Roy
0

Featured Post

Vote for the Most Valuable Expert

It’s time to recognize experts that go above and beyond with helpful solutions and engagement on site. Choose from the top experts in the Hall of Fame or on the right rail of your favorite topic page. Look for the blue “Nominate” button on their profile to vote.

  • 5
  • 4
  • 2
  • +6
Tackle projects and never again get stuck behind a technical roadblock.
Join Now