?
Solved

Best raid configuration for VMware server

Posted on 2009-04-28
15
Medium Priority
?
3,591 Views
Last Modified: 2012-08-13
Hi there,

I ordered a new server with a Xeon L5420 and 8 320Gb SATA disks on a 3ware 9650 raid controller and 12Gb of RAM (DDR3 ECC). I want to use the server to run multiple servers under vmware ESXi. To begin with it will run windows 2008 web server, one small freebsd server, one bigger one and one with heavy mysql usage.
My question is: How do I configure the raid system, so it is fast, reliable and still have the most diskusage?

Thanks,

Roger
0
Comment
Question by:rogerberger
  • 4
  • 3
  • 2
  • +3
14 Comments
 
LVL 22

Expert Comment

by:65td
ID: 24251386
The maximum VMFS partition is 2 TB.
So if one used raid 5 with 8 disks that gives 2.240 TB (>2 TB)
Raid 5 with 7 disks 1.920 TB.

How is 3i going to run off of a usb key or installed?
0
 

Author Comment

by:rogerberger
ID: 24251524
It's going to run from USB (need to add manually the driver for he 3ware card).
Raid5 with 7 disks would be fine, so the eight disk could be spare disk. But is RAID5 fast enough or muchslower than f.e. raid10?
0
 
LVL 4

Assisted Solution

by:Maasdriel
Maasdriel earned 60 total points
ID: 24251560
This question isn't really related to VMWare. It depends on what your needs are.
http://en.wikipedia.org/wiki/Redundant_array_of_independent_disks
Look at link above for the difference in RAID configurations.

The most common RAID configurations are RAID5 and RAID1.
Economy wise RAID5 you have the most usable diskspace.

For example if your MySQL database has a alot read I/O a RAID1 set will preform better then a RAID5 set. Downside is you will need more disks compared to RAID5.
0
Efficient way to get backups off site to Azure

This user guide provides instructions on how to deploy and configure both a StoneFly Scale Out NAS Enterprise Cloud Drive virtual machine and Veeam Cloud Connect in the Microsoft Azure Cloud.

 
LVL 22

Assisted Solution

by:65td
65td earned 60 total points
ID: 24251611
Does the Array controller have read / write cache on it?

Read performance should be as noted above similar for Raid 1 and 5, raid 1 has less I/O for writes than raid 5.

Can the array controller support a hot spare?

Or build it with a raid 1 of 320 and a raid 5 will 6 drives (1.6 TB).
0
 

Author Comment

by:rogerberger
ID: 24251773
The controller has read/write cache and has hot spare support.
Is a raid 1 system faster than raid5? Then another question: Is raid 5 with 8 harddisks faster than with 4 harddisks.
I was thinking about using 2 parts: 4 harddisks in Raid10 for the high I/O servers and the other 4 servers in raid5.  
0
 
LVL 8

Expert Comment

by:markzz
ID: 24252732
G-day Roger
If you were using SATA disk then a raid 5 or possibly a raid 50 would be fine but as you are using SATA I wouldn't expect great performance from the disks.
I've had a read about the 3Ware 9690SA it's not a bad piece of kit, it will alow you to create a single raid 5 and carve it into logical volumes of what ever size you want, I would suggest a 1TB max logical volume..
BUT
Yes there's the BUT.
As you are using SATA disks I'd likely use raid1 mirrors, meaning you will end up with 4x 320GB LUNS.
The issue with SATA is still the old IDE type IO management. Single IO Single responce. Sure they have loads of memory on disk (RAM) and it's now serial but it's all relative to the technology of the day and SAS is the common technology for server work loads.
You mentioned a MYSQL Server with high IO requirements, do you know how much IO is expected??
The idea of more spindles equates to shared IO load in large stripes writes is quite true but I would be a little cautious as the Dsisks are SATA.
I'd say raid 5 is out of the question.
Raid6 is possible but your IO write rates will be low.
Maybe a raid50 where you have 2x raid5's, so your first set of disk would be a raid5 of 4 disks this is then mirrored onto another set of 4 disks in a raid 5.. Hmm thinking about capacities but then of your 8 disks you only get the use of 3x 320GB. That's worse than using 4x raid1 mirrors. but it does give you 4 spindles rather than 2 to average the IO load over..
Maybe it's woth looking at the raid6 option, All 8 disks in a raide6, you'll get 6 disks useable with 2 parity disks. This would equate to 1.9TB of disk in a single LUN or you could split this up iinto smaller LUNs.
Raid 6 is a performance compromise between a raid1 and a raid5.
Be sure to read up on the controllers start sector offset and recommended Strip size remembering that VMWare ESX will write in .5MB, 1MB, 2MB or 4MB clusters depending on the size you set when creating your LUNs.
 This can make a significant improvement to your disk performance particularly with SATA disks.




 
0
 
LVL 8

Accepted Solution

by:
markzz earned 135 total points
ID: 24252753
Just to clarify
My choices would be either a raid6 or if you feel you need more performance than a raid6 can offer then it would be 4x raid1's
Again read up on stripe size and starting sector offsets. This is quite important.
0
 
LVL 56

Expert Comment

by:andyalder
ID: 24252968
Heavy mysql on SATA disks? Can you confirm what the disks are and more importantly the duty-cycle.
0
 

Author Comment

by:rogerberger
ID: 24254196
The machine is a supermicro 1026T-URF and th disks will be Seagate ST9320421AS, 320Gb, 7200RPM 2,5"
The raid controller is the 3ware 9650SE and not the 9690 which means SAS is not an option.
0
 
LVL 56

Assisted Solution

by:andyalder
andyalder earned 60 total points
ID: 24254972
Momentus 7200.3? I thought they were designed for laptops, I have no idea what the duty-cycle is for them, not posted in Seagate's specs. I very much doubt it is 100%.
0
 

Assisted Solution

by:Stachal
Stachal earned 60 total points
ID: 24256096
7200 is going to Hurt, With the current config i would def go with raid5 7 disks with a hot 8th spare. unless you want to cut your usable in half and go with a raid10 4/4. With the IO your SQl is going to produce you may have to go raid10 .  anyway you look at it 7200 sata disks are goign to slow you down, is this a production box?  i
0
 

Author Comment

by:rogerberger
ID: 24257749
Yes it's a production box. I experimented with a 1025 which is running 4 disks in raid5 and 4 disks in raid 10. It is running flawlessly for 8 months now. The problem is that this server will run vmware and more servers than the test server.
I think I can conclude that raid 10 is the best solution. I will lose some diskusage, but will have maximum I/O.

Oh, I use the 2,5" disks because the server has to be very green. The 1025 server with 2 XEON L5320 procs, 8 2.5" disks and 8Gb ram only uses 185W on full CPU usage. This server will probably use less.
0
 
LVL 8

Expert Comment

by:markzz
ID: 24472691
Yes I do object.
The question was actively answered by a number of contributors including myself.
I see that rogerberge has chosen to use a raid10 or raid1 volume configuration which was suggested by a number of contributors.
Sorry rogerberger, your question was given attention, discussion and answers from a few people and therefore deserves points assigned.
0
 

Expert Comment

by:Stachal
ID: 24479670
Also Remember to take into account ESXi only supports up to 2tb local VMFS storage patitions, you can create multiple vmfs patitions at 2tb.
0

Featured Post

Free Tool: IP Lookup

Get more info about an IP address or domain name, such as organization, abuse contacts and geolocation.

One of a set of tools we are providing to everyone as a way of saying thank you for being a part of the community.

Question has a verified solution.

If you are experiencing a similar issue, please ask a related question

Giving access to ESXi shell console is always an issue for IT departments to other Teams, or Projects. We need to find a way so that teams can use ESXTOP for their POCs, or tests without giving them the access to ESXi host shell console with a root …
How much do you know about the future of data centers? If you're like 50% of organizations, then it's probably not enough. Read on to get up to speed on this emerging field.
Teach the user how to rename, unmount, delete and upgrade VMFS datastores. Open vSphere Web Client: Rename VMFS and NFS datastores: Upgrade VMFS-3 volume to VMFS-5: Unmount VMFS datastore: Delete a VMFS datastore:
Teach the user how to use configure the vCenter Server storage filters Open vSphere Web Client:  Navigate to vCenter Server Advanced Settings: Add the four vCenter Server storage filters: Review the advanced settings: Modify the values of the four v…
Suggested Courses

809 members asked questions and received personalized solutions in the past 7 days.

Join the community of 500,000 technology professionals and ask your questions.

Join & Ask a Question