Simple Hyper V processor allocation explanation please!

Durges
Durges used Ask the Experts™
on
So the more I read about this the more frustrating I find it. It seems the more simply people try to explain it the more convoluted the explanations get.

Here is a simple example of a scenario.

Server 2016 Xeon 2 CPU's 8 Cores each
So this is 16 logical processors

I have 4 Hyper V VM's
2 x DC these need a small amount of CPU but it cannot be allowed to fall below a certain threshold say 20% as it serves DNC / DHCP etc
1 x File server
1 x RDS with 5-10 users max

64GB RAM

The server should be fine for the small amount of users and throughput, it is some basic files and there are no heavy apps in RDS.

All the servers should be able to take however much free resources they want.

How many VP should I assign per server.

My reading leads my to believe give them all 2 VP's and set max  / min limits to protect other servers and HV will share out free CPU time efficiently. My experience is that this makes them sluggish and the thresholds do not allow for high overall system resource. I would expect RDS to need the majority most of the time but if I am running WSUS ora backup process overnight when RDS is not being used it should give up its resources.

Any help appreciated, also how do I know HV is using all LP's not queuing time for the same ones from different servers.
Comment
Watch Question

Do more with

Expert Office
EXPERT OFFICE® is a registered trademark of EXPERTS EXCHANGE®
Andrew Hancock (VMware vExpert / EE Fellow)VMware and Virtualization Consultant
Fellow 2018
Expert of the Year 2017

Commented:
Okay, it can be confusing...

did you read fellow Expert Philip Elders' article on Hyper-V ?

https://www.experts-exchange.com/articles/13256/Some-Hyper-V-Hardware-and-Software-Best-Practices.html

https://www.experts-exchange.com/articles/31442/Practical-Hyper-V-Performance-Expectations.html

which explains Processors and Cores ?

Author

Commented:
So the gist of this I should be setting the DC's and file server to 2vCPU and the RDS to 4vCPU. When I do this the virtual machine limit in resources is very low, does this not mean it is limited to the amount of the host OS it can use, this is only in test at the moment so has no users but is very slow.
Philip ElderTechnical Architect - HA/Compute/Storage

Commented:
Why 2 DCs on a single host? Have a known good backup on the one good DC and that's not needed.

What is the disk setup? RAID? # Disks? # Arrays?
JavaScript Best Practices

Save hours in development time and avoid common mistakes by learning the best practices to use for JavaScript.

Author

Commented:
Thanks for the response, I appreciate the time from you guys.

RAID 1 SSD OS RAID 5 / 5 SAS SWAP Filestore, RDS Filestore. I know SSD is better for databases and the OS's especially the host don't require it. As I am currently the only user on the server IOPS isn't the issue.

The second DC will move when this goes prod. It will be replaced by a 2012 LOB app server which will have very low use. I was just trying to present it as is in a simple environment.

My whole question is what is a simple rule of thumb, I get when you have high performance specific scenarios like RDS farms or SQL you might need specific configurations. I realise rule of thumb isn't the perfect optimal setup. I can look at it and tune later. If the answer is there is no simple rule of thumb for a reasonable config then so be it. I have some relatively light use servers and want to reasonably use the host resources to whichever VM is being used at the time, this is a small headcount and usage varies greatly across server at different times.

Is my CPU count of 2/2/2/4

a) Using all logical processors
b) Using all or most total available resources or only 12 of the 16
Philip ElderTechnical Architect - HA/Compute/Storage
Commented:
Um huh? Why close before any of us get a chance to respond? The reason was not very polite IMNSHO either.
Technology and Business Process Advisor
Most Valuable Expert 2013
Commented:
As a rule of thumb, I set ALL VMs to 2 vCPU for almost all workloads.  At which point, you MONITOR the VMs and if they appear sluggish you find out why.  Is it because you assigned too little RAM?  Slow disk?  CPU usage?  Performance Monitor, Resource Monitor, and Task MAnager can all give you insight into this.  If it's disk, then you have to potentially spend a lot to get a faster disk subsystem.  If it's CPU, you can add another vCPU (I recommend ONE AT A TIME).  If it's RAM, increase the RAM.  Sluggish performance is not, in and of itself, indicative of a CPU resource issue.  

Further, you start with two and go up slowly as needed because of the way the hypervisor allocates processing time to the VM. Philip has an article on this and there's another article that explains it here:
https://www.zdnet.com/article/virtual-cpus-the-overprovisioning-penalty-of-vcpu-to-pcpu-ratios/

Author

Commented:
Thanks Lee, this is what I was looking for, sorry Philip I was rude and I apologise, your guide was great I was just frustrated and looking for a simple jumping off point.

Author

Commented:
Thanks everyone

Do more with

Expert Office
Submit tech questions to Ask the Experts™ at any time to receive solutions, advice, and new ideas from leading industry professionals.

Start 7-Day Free Trial