• Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 912
  • Last Modified:

What's the caveats in using VMFS extends on top of RAID-5 ?

Hi All,

I've got Dell Server with 6x SATA 450 GB HDD configured with RAID-5
so approximately the total capacity is (6x450)-450 = 2.25 TB

for the above datastore I want to use as one single datastore by using multiple extends, can anyone suggest me is there any caveats or performance penalty in using one large datastore on top of RAID-5 disks ?

Thanks,

JJ
0
jjoz
Asked:
jjoz
  • 7
  • 3
  • 2
  • +3
6 Solutions
 
arweeksCommented:
Raid 5 has performance issues for writes.  Also, on the off chance that you get an array failure it's going to be a hell of a restore window.

That being said, how importance is performance? What you suggest is certainly flexible, easy and gives you the best storage capacity.
0
 
jjozAuthor Commented:
the purpose of this datastore is to server as the backup repository of all the SAN VMs image so i wonder what would happens if one of the disk got failed, would I lose the whole datastore ?
0
 
jakethecatukCommented:
a lot will depend on what you're VM's are doing.

also, if you lose the first extent, then you have lost them all.

you will need to make sure that you format the VMFS correctly with the 8mb block size.

of course, you will only actually get around 2.15tb in total.  
0
Improve Your Query Performance Tuning

In this FREE six-day email course, you'll learn from Janis Griffin, Database Performance Evangelist. She'll teach 12 steps that you can use to optimize your queries as much as possible and see measurable results in your work. Get started today!

 
arweeksCommented:
If one disk fails, you're fine - that's the idea of raid 5.  You'd only lose the lot if you lost two disks before the array had rebuilt onto a new drive.
0
 
jjozAuthor Commented:
wow that's scary, is there any way to get the data back if something happens on the first extends ?
0
 
coolsport00Commented:
Performance will be ok, depending on what you're using the RAID5 for. I use RAID5 with everything and it works fine (Exchg, SQL, and all VMs). RAID10 is your best performance RAID, but you lose 2 disks. You don't lose anything having added extents because it's the same storage. You're still reading/writing to a RAID, so no caveats.

I always recommend splitting up storage though, to multiple datastores so as not to have "all your eggs (VMs) in 1 basket". You can separate your VMs across mulitple datastores so not all your infrastructure goes down in the event of multiple disk failure.

Regards,
~coolsport00
0
 
coolsport00Commented:
The only way to recover VMs is if you have a VM b/u solution, using Veeam, vRanger, or some tool like that. But that requires you to have a minimum of vSphere Essentials for the API to backup VMs.

~coolsport00
0
 
jjozAuthor Commented:
thanks for the suggestion coolsport, my plan is to have one Windows Server 2008R2 Vm to contains 2 TB VMDK which is being used as the backup container of all the other VM in my SAN, I'd like to know what might be the problem in terms of performance and reliability if there is any.

seems that I involve so many layer in here:
RAID-Layer
VMFS Layer + extends
VMDK
NTFS inside VM
Backup VM files
then files inside the VM
0
 
coolsport00Commented:
Well...RAID is the "backbone" solely. VMFS is just the file format and everything else are just files (minus the NTFS partition file format of 1 VMDK, but even then it's still just a file [vmdk] with NTFS  formatting). If you're concerned more about performance, rather than the amt of storage you have, you should consider RAID10.

~coolsport00
0
 
DavidPresidentCommented:
Easy - first generalize the answer ....  The plus/minus of RAID5 is the same no matter what O/S you use, so answer it for any O/S you are familiar with.

With extends, your partitions are not contiguous, so there is an additional performance hit when an app doing sequential I/O needs to do another seek to get to where the rest of the data is.  As VMWARE is effectively a random I/O environment from perspective of the storage, it won't be that much of a difference.
0
 
bgoeringCommented:
You usable size for 6x450GB in RAID5 is 2095.4 GB which exceeds VMware's 2TB limit for a LUN. So you are correct in that you will need to split your array into at least 2 LUNs. I have a similar situation with a Dell R710 with 6x500GB SATA drives, but I chose to configure one as a hot spare and the remaining drives fit under the 2TB limit so I did not configure multiple extents for local storage.

Everyone is correct that RAID 10 is a bit faster, but like coolsport00 all of my ESX/ESXi storage (both SAN and local) is configured as RAID5 (except for database log files) and my performance is acceptable. Read performance is as good as (and often better) for RAID5 because there are more spindles to stripe the data across. In your case with 6 drives you will have 5 spindles of data for reading (the other for parity) for each stripe, while in a RAID 10 setup you will only have 3. The penalty is incurred for high write activity workloads such as database logging. The short answer is the best RAID level selection is highly dependent on your expected workload.

One caveat is that for whatever RAID you choose, make sure your RAID controller is equipped with battery backed write cache, and is configured for write-back (as opposed to write-through) write caching.

Hope this helps
0
 
jjozAuthor Commented:
Thanks Mr. bgoering,

so in this case
RAID - 5 is for the application that requires good read and not bothering the write performance
RAID - 10 is for the application that requires both good read and write performance
0
 
DavidPresidentCommented:
As a generalization, above is correct.
If your RAID10 is incorrectly configured, grossly incorrectly configured, then RAID5 could outperform it.  Same is true with any RAID level.  Remember I/Os per second is mutually exclusive to throughput.

A pair of RAID1s will deliver more IOPs then a single RAID10, provided the I/Os are hitting both of the RAID sets equally.  Also, in a VMWARE environment, you are more likely to have highly random I/O (when eventually seen by the physical disk drives), so you should try to optimize IOPs with a higher priority than sequential throughput, where RAID10 is best.


0
 
jjozAuthor Commented:
thanks for all of th explanation guys,
so now I understand the concept behind RAID :-)

the attached picture might be a "de javu" for some of you here since i got it also for my other thread, however after i read this topic the Random IO is the most important thing here for VMware.
iSCSI-BlockIO-WriteThru.jpg
0
 
jjozAuthor Commented:
the above picture is showing the benchmark through the iSCSI to the Openfiler VM on my VMware ESXi.

i ran the DIskSPeed on my physical Windows server 2003 which connects using iSCSI to the OpenFiler 2.3
0
 
jjozAuthor Commented:
thanks guys for the explanation !
0
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

Join & Write a Comment

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.

  • 7
  • 3
  • 2
  • +3
Tackle projects and never again get stuck behind a technical roadblock.
Join Now