Peter Wilson
asked on
What is the best storage configuration for file share server
Hi experts,
what is the best storage configuration for a VMware virtualized file server in a local storage host.
I have 3 TB of storage on an old single file server (physical) so I was planning on splitting up the data into like-minded shared clusters of smaller hdds, e.g. 500GB, 700GB, 500GB, 800GB, 200GB & 300GB. The 3TB includes growth factor as well.
server 2012r2
esxi5.5
vcenter5.5
Thx.
what is the best storage configuration for a VMware virtualized file server in a local storage host.
I have 3 TB of storage on an old single file server (physical) so I was planning on splitting up the data into like-minded shared clusters of smaller hdds, e.g. 500GB, 700GB, 500GB, 800GB, 200GB & 300GB. The 3TB includes growth factor as well.
server 2012r2
esxi5.5
vcenter5.5
Thx.
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
more physical disks (spindles) outperform 1 physical disk. Partitioning just makes it worse as it increases the seek times.
Back in the day when motherboards only supported 2-4 drives and would only boot off of 1 drive and you wanted to keep items separated then partitioning. These days most motherboards support 4 or more sata drives and with the advent of USB 3.0/3.1 one can easily have 10 drives. If you then split the reading/writing among these 10 drives, each drive will be able to sustain 100 iops giving you an effective throughput of 1,000 iops.
You have to look at capacity and performance. 1 SATA/SAS15 drive will give you 100-150 IOPS maximum.
The most up to date advice is to not partition drives but add additional drives
Back in the day when motherboards only supported 2-4 drives and would only boot off of 1 drive and you wanted to keep items separated then partitioning. These days most motherboards support 4 or more sata drives and with the advent of USB 3.0/3.1 one can easily have 10 drives. If you then split the reading/writing among these 10 drives, each drive will be able to sustain 100 iops giving you an effective throughput of 1,000 iops.
You have to look at capacity and performance. 1 SATA/SAS15 drive will give you 100-150 IOPS maximum.
The most up to date advice is to not partition drives but add additional drives
Depends on your requirements, if you are building a 3TB file server, why do you want to create lots of disks ?
ASKER
For higher iops, I thought it was a better practice and consensus based on my last question regarding this topic (https://www.experts-exchange.com/questions/28956827/Best-practice-for-VM-drive-sizes.html), and flexibility.
and where are you going to be storing all your individual virtual disks, on the same VMFS datastore ?
e.g.
disk0 - OS
all these disks data
disk1 - 500GB
disk2 - 700GB
disk3 - 500GB
disk4 - 800GB
disk5 - 200GB
disk6 - 300GB
all on datastore1
OR
disk0 - System OS
data disks
disk1 - 3TB
on datastore1
Test it and see if the IOPS are different, you can quickly create a single disk, and two disks, and use CrystalDisk Benchmark to quickly test, I think you will find the IOPS will be similar!
and what is the underlying VMFS datastore, based on number of disks and RAID type ?
e.g.
disk0 - OS
all these disks data
disk1 - 500GB
disk2 - 700GB
disk3 - 500GB
disk4 - 800GB
disk5 - 200GB
disk6 - 300GB
all on datastore1
OR
disk0 - System OS
data disks
disk1 - 3TB
on datastore1
Test it and see if the IOPS are different, you can quickly create a single disk, and two disks, and use CrystalDisk Benchmark to quickly test, I think you will find the IOPS will be similar!
and what is the underlying VMFS datastore, based on number of disks and RAID type ?
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
@Andrew - Yes all disks on same datastore (vmfs 5), which is on fixed storage within the host, raid5 10 15k rpm spinners.
@David - I understand now...so in a virtual environment iops is only taken directly from the hypervisor...whatever it supplies is what is used. makes sense. thanks for that clarification!
@David - I understand now...so in a virtual environment iops is only taken directly from the hypervisor...whatever it supplies is what is used. makes sense. thanks for that clarification!
so therefore, all the virtual disks, are going to have roughly the same IOPS.
VMFS 5 datastore - IOPS per virtual disk = RAID 10 - 10 disks at 15rpm
Theoretical IOPS per datastore = 1026, so 1026 max per virtual disk....
and to be honest with you RAID 5, is no longer recommended as enterprise suitable, as you only have 1 disk failure in your 10 disks.
VMFS 5 datastore - IOPS per virtual disk = RAID 10 - 10 disks at 15rpm
Theoretical IOPS per datastore = 1026, so 1026 max per virtual disk....
and to be honest with you RAID 5, is no longer recommended as enterprise suitable, as you only have 1 disk failure in your 10 disks.
ASKER
yes I'm aware raid5 is no longer supported - we plan to go to raid10 in our new hypervisor. so hypervisor aside, and talking specifically about vm virtual hdd configuration...is there any advantage to multiple virtual hdds vs 1 single big GPT virtual disk. you had mentioned chkdsk and longer virus scans...e.g. more data at analyze for any give reason. all this said, what would you recommend I setup for a share server: a. multiple smaller virtual disks or 1 large virtual disk for win 2012r2?
simple administration, single hard drive
ASKER