solution for Hyper-v + replication + backup

[leegcvcc]

Hi all,

I'm long overdue a refresh of a 10 year old HP DL380 gen6 3 host (almost diskless) and fibre channel MSA2024 SAN infrastructure using ESXi. This was installed by a 3rd party so I didn't design or scope it.

 

We have a Microsoft OVS EES agreement and datacenter editions for 32 cores, so I need to look at Hyper-v to cut costs. We are a 1000 user school so there are probably 350 concurrent users at peak times. We have about 7tb of used data across all 30 VM's, no dedupe so the 7tb is a count of GB across all VM's that are fully backed up with vRanger (5 VM's are spiceworks, oh for multiple instances per server!)

 

Having looked at refresh options, it seems that I don't need to replicate what we have in terms of a SAN solution. I am sick of worrying about a SAN failure even though dual PSUs and Controllers etc, as backup restores would take an age! We will be moving more services to the cloud over the coming years but still need to cater for our existing storage requirements in the meantime.

 

I would like to use Dell servers and Veeam Backup and replication (unless persuaded otherwise).

 

We run 1 full SQL 2017 and about 4 SQL Express VMs for various systems

 

 

I am after some opinions on the following please as I had read articles for England on the many ways to 'skin a cat' and wondered if the following is feasible and will work in our environment.

 

 

2 x identical 16 bay Hosts (maybe a refurb Dell R730xd or 740? etc)

 

each would have:

 

2 x Intel 8 core CPUs (could be more cores but isn't licensing increased after a 16 core usage?)

512gb ram

8 x 2tb SSD in raid10 array (8tb usable)

8 x 2tb HDD in raid 5 array (approx 10tb usable)

 

Dual PERC controllers ( 1 for each array?)

4 x gigabit nics

4 x 10gbe nics

 

I would install the full 2019 datacenter Server with Hyper-v role on Host-A and Host-B (i can't be setting up the hypervisor without the windows tools as I've just had a look at Hyper-V for the first time this week so it's a bit much to manage without a GUI)

 

The hyper-v would be on the RAID 10 SSD arrays

 

What I'd like to do is have half the VMs on the SSD array on Host-1 and the other half on the SSD array on Host-2 and then replicate Host-A to Host-B HDD array and then replicate Host-B SSD to Host-A HDD using Veeam.

Having used vRanger I am not familiar with Veeam either so Hyper-v and Veeam is totally new ground.

 

The NICs all go back to a gigabit Cisco 3750 routing stack. I would dedicate 3 gigabyte NICs to production and one Nic for Management. I could connect the Hosts together using direct 10gbe NICs as they are in the same rack

 

Is it possible to do all of this?

 

I assume I would need to make each host's HDD arrays available to Veeam. Would this be using SMB?

 

I assume it would be advantageous to pack some HDDs and use the DL380s to install Veeam on so it processes the jobs etc and not on the hosts?

 

What sort of replication intervals could be achieved in this set up? if it's even feasible?

 

How quickly could failed VM's be brought back up on the running host after a failure on other host? If feasible?

 

I could also use a DL380 as a backup repository for Veeam and attach an existing LTO4 tape drive for Disk to Tape and leaving the hosts free from this too. I also have 2 x 4 CPU R820 512gb RAM refurbs which I can use but licensing would increase hugely I think.

 

I have approx 30k budget (excluding Microsoft ov-S) so need to get something up to the job ideally for 5 years. As I said, I can't ignore refurbs although not had any experience with them.

 

I have looked at Hyper-v clustering and Starwinds hyperconvergence but they all seem to add a layer of complexity and extra consideration which I am sick of. Other systems seem complex enough without the backend adding to it if unneccesary

 

Am i dreaming?

[Jcx500]

i would use storage spaces direct for this soultion

[leegcvcc]

Cheers for reply

 

Why would you?

[StoneGroup_JackW]

Hey @leegcvcc we can definitely help with that with our equinox solution. can you please drop me a PM with some contact information and i can send you some details over? we also have lots of case studies and schools who have this solution

 

Kind regards

[Reboot_IT]

We use Storage Spaces Direct with 3 hosts. Been using 2 instances of this solution for 18 months and am very happy with it. Server core is installed on them all with the hyper-V role. We use DPM for backups.

[mavhc]

AIUI the advantage of SSDirect is your VMs are effectively cloning themselves continuously to the other servers, so if any one fails the remaining ones can continue to run the VMs. 0 downtime when it's working properly

 

And the disadvantage is complexity, requiring specific hw, etc

 

You can always replicate your VMs to other hosts and manually start them if the other host fails. Depends if you want 99.999% uptime or only 99.99%