ATS 5.1.1 And ESXI Ubuntu Guest

[Jason Strongman <ja...@gmail.com>]

Hypervisor Version - ESXi 5.5
Guest OS - Ubuntu 12.04.5 and 14.04.1 (64bit generic kernel)


While testing ATS in our Ubuntu guest VMs we noticed a significant
performance difference(for the worse) between another RHEL 6.5 VM. Both VMs
run in the same hypervisor and both VMs have the same 'hardware' specs.
Both VMs also received the same identical 5gbps Spirent performance tests.
Each test ensured that the tested objects are served from RAM cache. So it
looks like(just guessing) the RHEL kernel may be optimized for virtual
environments in terms of memory management??
I do notice the library and compiler versions differ between the ATS
instances, the compile time options are the same. So I am not sure if that
may be prove to be an issue as well.

That said, are there any OS tuneables that I should review  within my
Ubuntu builds?
Next I am going to try the various Ubuntu kernels and see if that results
in the same RHEL performance.

Thanks!

Re: ATS 5.1.1 And ESXI Ubuntu Guest

[Jason Strongman <ja...@gmail.com>]

Jason.. thanks for your response.

Yeah I noticed vmware specific driver versions differ between RHEL 6.5 and
Ubuntu Precise/Trusty.. With the versions on RHEL 6.5 being a little behind
latest generic kernel of Precise/Trusty

Next, for my precise build, I simply installed 3.2.0-29-virtual, then
installed all the vmware provided modules(see below) built against this
kernel.

http://packages.vmware.com/tools/esx/5.5u2/ubuntu/dists/precise/main/binary-amd64/index.html

Performance increased significantly, but RHEL still won out. Right now I am
looking at tweaking OS tuneables and see how much mileage that gets me.



On Thu, Nov 20, 2014 at 4:48 PM, Jason Giedymin <ja...@gmail.com>
wrote:

> I would record the kernel, and lspci, dmesg the hw/driver versions. You
> could be using different driver versions between the kernels.
> After that I’d look at the sysctl diffs between the two systems.
> Make sure the esxi resource entitlements and priorities are the same for
> each guest (which I’m sure you’ve already done, doesn’t hurt to double
> check).
>
> Wondering if your using my drivers for ESXi (though I’m sure they’d be out
> of date by now, and depends on what your networking hardware is).
>
> -J
>
> > On Nov 20, 2014, at 5:33 PM, Jason Strongman <
> jasonstrongman2016@gmail.com> wrote:
> >
> >
> > Hypervisor Version - ESXi 5.5
> > Guest OS - Ubuntu 12.04.5 and 14.04.1 (64bit generic kernel)
> >
> >
> > While testing ATS in our Ubuntu guest VMs we noticed a significant
> performance difference(for the worse) between another RHEL 6.5 VM. Both VMs
> run in the same hypervisor and both VMs have the same 'hardware' specs.
> Both VMs also received the same identical 5gbps Spirent performance tests.
> Each test ensured that the tested objects are served from RAM cache. So it
> looks like(just guessing) the RHEL kernel may be optimized for virtual
> environments in terms of memory management??
> > I do notice the library and compiler versions differ between the ATS
> instances, the compile time options are the same. So I am not sure if that
> may be prove to be an issue as well.
> >
> > That said, are there any OS tuneables that I should review  within my
> Ubuntu builds?
> > Next I am going to try the various Ubuntu kernels and see if that
> results in the same RHEL performance.
> >
> > Thanks!
> >
> >
>
>

Re: ATS 5.1.1 And ESXI Ubuntu Guest

[Jason Giedymin <ja...@gmail.com>]

I would record the kernel, and lspci, dmesg the hw/driver versions. You could be using different driver versions between the kernels.
After that I’d look at the sysctl diffs between the two systems.
Make sure the esxi resource entitlements and priorities are the same for each guest (which I’m sure you’ve already done, doesn’t hurt to double check).

Wondering if your using my drivers for ESXi (though I’m sure they’d be out of date by now, and depends on what your networking hardware is).

-J

> On Nov 20, 2014, at 5:33 PM, Jason Strongman <ja...@gmail.com> wrote:
> 
> 
> Hypervisor Version - ESXi 5.5
> Guest OS - Ubuntu 12.04.5 and 14.04.1 (64bit generic kernel)
> 
> 
> While testing ATS in our Ubuntu guest VMs we noticed a significant performance difference(for the worse) between another RHEL 6.5 VM. Both VMs run in the same hypervisor and both VMs have the same 'hardware' specs. Both VMs also received the same identical 5gbps Spirent performance tests. Each test ensured that the tested objects are served from RAM cache. So it looks like(just guessing) the RHEL kernel may be optimized for virtual environments in terms of memory management??
> I do notice the library and compiler versions differ between the ATS instances, the compile time options are the same. So I am not sure if that may be prove to be an issue as well.
> 
> That said, are there any OS tuneables that I should review  within my Ubuntu builds? 
> Next I am going to try the various Ubuntu kernels and see if that results in the same RHEL performance.
> 
> Thanks!
>    
>