0

I have an HP ProLiant DL380p Gen8 (VMWare 7u3 installed) with 2 pieces of 960GB SSD RAID1 arrays (total of ~1+1=2GB usage). The rest of the configuration is like

  • 24 CPUs x Intel(R) Xeon(R) CPU E5-2697 v2 @ 2.70GHz (2x12 Cores)
  • 256GB RAM

The issue I am having is with the disk performance. Today I switched off one of my Ubuntu VMs and started moving its folder from one datastore to the other (I mentioned that I have 2 pieces of 1TB RAID1 arrays and each is a seperate datastore). Then I had a look at the Monitor page of VMWare and saw a chart like that for disk performance.

enter image description here

You see the spikes here? Why doesn't it continue moving the VM folder with a steady speed? I don't understand if there is a bottleneck on that box or smth. Interesting thing is, it's always showing a similar chart for disk performance whenever I try to do smth with disk IO.

During the folder moving process the CPU and RAM usages were similar to this all the time (That is a development server and not much going on on other VMs).

enter image description here

Can you please help me understand why I cannot leverage the full potential of SSDs (530MB Read / 480MB Write speeds we are talking about here)

Thanks!

9
  • 1
    ESXi? Storage performance is all about the storage controller and the drives. What type of controller and SSDs are you using on the DL380G8? The P420i should be decent.
    – Zac67
    Apr 27, 2022 at 9:21
  • Yes, VMWare ESXi 7 latest version is installed on the box. And yes, there is a P420i controller inside it. Disks are Samsung 870 EVO.
    – ilter
    Apr 27, 2022 at 10:11
  • The 870 EVO are consumer-grade but should do (unless the duty cycle kills them). I'm assuming the P420i has FBWC. The "moving its folder" is about guest files? Then the bootleneck might be the guest file system (e.g. many small files) or processing.
    – Zac67
    Apr 27, 2022 at 10:26
  • What I did basically is that I opened the datastore1 (from VMware ESXi web console) and choose the VM folder and clicked on Move button and chose datastore2, after shutting down the guest OS.
    – ilter
    Apr 27, 2022 at 10:30
  • You are aware that this might render the VM unstartable? You should storage migrate the VM instead. Direct datastore operations can be (painfully) slow at times.
    – Zac67
    Apr 27, 2022 at 10:33

1 Answer 1

1

The DL380p g8 with the P420i and SSD perform badly sadly. For more details please see;

Please see that answer; https://serverfault.com/a/732298/256639

By default, the SmartArray disables the disk's private DRAM cache, significantly lowering SSDs performance.

You can try to reenable it, but you had to be sure (by mean of testing) that this will not led to data corruption in case of power loss.

or from another thread;

After additional testing, HP's SSD's don't perform any better. I get roughly the same results. That said, make sure your write cache is set to 100% write, that improves the results noticeably. Also consider enabling drive write cache as this helps the consumer SSD's perform better.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .