Jump to content

Recommended Posts

Posted

Just thinking forward to when I need to replace our Hyper-V server. The VMs running on that server are DC vm and Data vm. The Data vm doesnt have a great deal installed as most of our data is in the cloud, including our MIS.

 

In the past I have always assigned all the drives to a single raid 10 array, with a hot spare.

 

I was thinking about separating out the OS © drives from the Data Drives (D). Is this a good thing for a Hyper-V server? I assume it is, but wanted to check.

 

I was either thinking of:

a)OS © on 2 drives in RAID1 + hot spare

 

or

 

b)OS © on 4 drives in Raid 10 + hot spare

 

The Data (D) drive will be x amount of disks in Raid 10 + hot spare, for the VMs to run from

 

Is there a preference for a or b setup?

Posted
While our implementation is different (using a SAN for the drives, connected via fibre to the VM hosts), we do separate our our OS and data drives. We have two pools of drives, 15k SAS drives or 7.2K MDL drives, with both sets using RAID5 (I think). Typically, all our OS drives are stored on the SAS drives, with the slower storage for data (or even the OS of our low-end VMs).
Posted
What sort of drives are you using, and how big are they?

Previously I have purchased 15k SAS drives, but thinking of getting SSD drives.

 

Have not sat down and worked out space requirements yet, but guessing the Hyper-v C drive needs to be 100gb, and the D drive for both the VMs to live, needs to be around 500gb total. Plus any hot spare drives

Posted

For local storage all drives in RAID 10. A partition for the host OS and a different partition for the Host Data (VMs)

 

What advantage are you getting from having the Host OS on a different RAID array?

 

If you have different Drive sizes or types then maybe?

Posted

For local storage all drives in RAID 10. A partition for the host OS and a different partition for the Host Data (VMs)What advantage are you getting from having the Host OS on a different RAID array?If you have different Drive sizes or types then maybe?

 

So just keep it the way I have always been doing it.

 

The reason I mentioned it, I seem to remember someone saying its best to separate out the partitions onto different arrays, but could be remembering that wrong.

 

Thanks

Posted

I cant see any reason to split them up these days.

If you had ssds and hard drives maybe you would put the host os on the hard drives so the vms have the faster storage. Possibly for some specific workload but in general one RAID 10 is fine for most workloads.

  • Thanks 1
Posted
The other option to consider, maybe, is using Storage Spaces instead of RAID volumes. That way, you could use tiered storage and use a mixture of SSDs and 15k spinning rust
  • Thanks 1
Posted

I went again with 2 x OS SSD in raid 1 with win installed and then X SSD in Raid 5 or X HDD in Raid 6 for data. This has proven over the last 20 years to be solidly reliable, very fast and extremely cost effective. We never ever had the budget for £50k cluster SAN systems anyway - but frankly from what I've seen, the more complexity you inject into storage systems to gain scalability, the slower and less reliable it becomes, or you can keep the speed and reliability but it becomes exorbitantly expensive.

 

Always wanted to move to something like Vmware Vsan when that first came out but the licenses were horrific and Storage spaces (a few years ago) looked a bit of an unreliable mess.

 

Not sure what I'll do for the next server refresh.

  • Thanks 1
Posted
The other option to consider, maybe, is using Storage Spaces instead of RAID volumes.

 

This is what we've gone for - you can buy servers with certification / warrenty that they work with Storage Spaces. Our configuration is two small-ish / fast-ish disks on a RAID1 controller for the OS, several larger (8TB, half a dozen I think), slower drives on an HBA card and two NVME disks (also on the HBA controller, if I remember correctly). You install Windows on the OS RAID1 drive and hand it the collection of large disks and NVME disks as a cache and leave it to figure out how best to utilise them.

 

A possible alternative is a similar set of disks as above but with the storage on a NAS. We have a Synology NAS, also with a bunch of slow, large disks and a couple of NVME disks, you put the drives in the device and leave it to configure them how it sees fit. You can then export the resultant high-capacity, large-cache storage volume via iSCISI, possibly via a dedicated 10GbE directly to your server. A suprisingly affordable option - around £1,000 for the physical NAS, plus whatever disks you want to populate it with, and possibly a touch more reliable / easy to set up than Windows Storage Spaces.

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...