vmtools – sick of argueing the benefits so lets splunk in.

Are you sick to death of asking your Windows Admins, Linux admins, and on occasion your security team to install vmtools.

I know I am.

So I thought once and for all I would benchmark the difference.

A couple of easy to read comparison charts before we start.

ConversionChart

The scenario.

ESX 4.1 build 8000380

Two Windows 2008 R2 (X64) servers sitting on the same esxhost ESX 4.1connected through a dvswitch.

Each Windows server is setup as follows.

1 cpu

8 Gb ram

Local Storage SAS 6GB

Host specs

Host used

Benchmark tool used

The trusty old iometer

Each test was 5 minutes per payload allowing for peaks and troughs.

Network – all figures are in mbps (unless specified otherwise)

Test 1 – No VMTOOLS – network e1000 mtu 1500 

vs 

VMTOOLS – network e1000 mtu 1500

e1000-comparision2

Test 2 – No VMTOOLS – network e1000 dvswitch 9000

vs

VMTOOLS – network e1000 dvswitch 9000

e1000-comparision3 - dvswitch 9000

Test 3 – No VMTOOLS – network e1000 – nic + dvswitch 9000

vs 

VMTOOLS – network e1000 – nic + dvswitch 9000

e1000-comparision3 - dvswitch 9000 - nic

Tests coming soon.

Test 4 – No VMTOOLS – network e1000 – nic + dvswitch 9000

vs 

VMTOOLS – network vmxnet3 – nic + dvswitch 9000

e1000-comparision4 -vmxnet3 - nic dvswitch 9000

Test 5 – No VMTOOLS – network e1000 – dvswitch 9000 (nic 1500)

vs 

VMTOOLS – network vmxnet3 – dvswitch 9000 (nic 1500)

e1000-comparision4 -vmxnet3 - dvswitch 9000

DISK (Read performance only)

Test 1 – No VMTools – Disk IO

vs

VMTools – Disk IOdisc - vmtools vs no vmtools

Test 2 – No VMTools – Disk IO

vs

VMTools – Disk IO Paravirtual

disc - vmtools para vs no vmtools

Conclusion.

Okay so I am a bit suprised by the results.

Most definalty vmxnet3 is the way to go if you have two servers that are chatty and can put them on the same host to utilise 10Gb through the vswitch or dvswitch. The graphs really sell that point.

But vmtools for disc performance, wow it almost seems like vmtools slows the thing down. However the boot up speed is quicker and general feel seems much quicker on windows 2008 R2 after switching to paravirtual scsi. The numbers do not lie though.

Keeping in mind……..

The big benefit of vmtools is that the ESX Host does not need to put an emulation layer between the guest and the device, so with vmtools installed we do save on resources on the ESX Host meaning we can run more on the host.

But if you like me use to get frustrated with admins about upgrading vmtools, we now know it really is not that important.

UNLESS….

You have start up and shutdown scripts. – not power off but guest shutdown.

Run RVTOOLS to gather information about your virtual environments – IP settings and partition info for example.

Like your ESX environment to run at its peak efficiency..

 

Advertisements