Dell PowerEdgeT20 - VM performance issues

tom01 · 23 Mar 2020 at 16:37

Hi Guys,

I'm having some problems with performance in some VMs on my T20. I've checked the CPU/Memory utilisation within ESXI and it looks absolutely fine.

The problem is I have especially on one Windows 10 (Fresh install) VM is that they are running very very slow, MSTSC sessions dropping, extreme lag etc. When I've checked the task manager on the affected VM CPU and Ram utilisation looks healthy, lots of head room with around only 27% Ram usage and 30% CPU.

I've allocated 2 CPU's to more intensive VM's, 1 CPU to the normal ones and all have a decent allocation of RAM.

Sometimes when I remote in to the box it runs absolutely fine but then other times it is completely unusable. I have 4 VM's running in total at the moment.

ESXI reports the following stats in monitor:
CPU:
Used: 4.5 GHz Free: 8.3 GHz

Memory:
Used: 22.66 GB Free: 9.25 GB

Spec:
4 CPUs x Intel(R) Xeon(R) CPU E3-1225 v3 @ 3.20GHz
32GB Ram

Sometimes when I remote in to a VM it runs very very slow for a while and then will become blistering quick, almost like there is some sort of lag whilst the hyper-visor re-allocates resource to the active VM. However I don't understand 1) Why this takes so long to get decent performance restored, 2) Why performance is so bad when there appears to be lots of computational resource available in the first place?

Has anyone else experienced this? Or can shed any light on what is happening?

Throrik · 23 Mar 2020 at 21:06

Are you seeing any spikes in ESXTOP? CPU Ready/Wait status etc?

Do the VMs have the latest VMware Tools?

#Chri5# · 23 Mar 2020 at 22:17

What do the performance metrics for the storage look like?

Avalon · 24 Mar 2020 at 13:41

Check the storage IO load, CPU/RAM usage isn't the only relevant metric.

tom01 · 24 Mar 2020 at 15:18

Avalon said:
Check the storage IO load, CPU/RAM usage isn't the only relevant metric.

Thanks for the reply, anything I should be looking out for in particular? I'm ok with CPU/Mem metrics but not IO loads.

Edit: Here are some stats (MB/s).

HDD spec:
1TB 7200rpm 32MB Cache - 6Gb/s

HDD with all my datastores on:
* Read Rate: Average: 0.19 // Max: 9.52 // Min: 0
* Write Rate: Average: 1.13 // Max: 11.21 // Min: 0.15

Total Disk usage stats for the specific VM: Average: 0.54 // Max: 9.73 // Min 0

#Chri5# · 24 Mar 2020 at 15:29

How many VMs have you got? Are they are all running from a single 1TB disk?

tom01 · 25 Mar 2020 at 09:13

I have 4 VM's at present, all datastores are on a single 1TB disk.

#Chri5# · 25 Mar 2020 at 15:36

What about the latency for the disk in ESXi?

tom01 · 25 Mar 2020 at 18:13

ESXI is installed on the same disk as the VM's.

#Chri5# · 26 Mar 2020 at 09:49

Sorry, I meant the disk latency in ESXi (should be where you found read/write rates). I wonder if the single disk is getting overwhelmed, so when the the VM goes slow, look at the latency. IIRC, you're limited to the last hour of metrics though.

andy_mk3 · 26 Mar 2020 at 10:46

tom01 said:
I have 4 VM's at present, all datastores are on a single 1TB disk.

That's your problem there right away. You need some solid state storage.

TheRealDax · 4 Apr 2020 at 23:10

Like "no idea what" said.
Use one Solid State for ESXi (or two for mirror), One Solid State for VM Operating Systems and applications, use the 1TB for your data (or also two in mirror if you want to be save not losing anything).