Last updated on September 4, 2016 by Dan Nanni
If you are using a virtualization hypervisor, one of your main concerns will be its performance, or in another word, its virtualization overhead. How much overhead is introduced by the virtualization layer will determine the raw performance of guest virtual machines (VMs) running on a hypervisor.
To compare virtualization overhead of two virtualization software packages, I created one VM on VMware Player, and another VM on VirtualBox, both of which have exactly the same specifications. Then I ran various PTS tests to obtain performance benchmarks from both VMs. Since two guest VMs have exactly the same hardware specification, the lower the virtualization overhead is, the better performance a VM will get.
I conducted the performance experiments on one VM at a time, so that there is no resource contention between the two VMs. That is, I powered on VMware Player VM, completed a series of tests on it, and powered off the VM. Next, I powered on VirtualBox VM, repeated the same set of tests with it, and powered it off.
I did performance tests in three categories: (1) disk, (2) processor and (3) memory. For each category, I randomly chose several tests among available ones from Phoronix Test Suite.
BlogBench: Final score. More is better. VMware Player won!
Dbench: MB/s. More is better. VMware Player won!
IOzone (Read): 64Kb record, 512MB file. MB/s. More is better. VirtualBox won!
IOzone (Write): 64Kb record, 512MB file. MB/s. More is better. VirtualBox won!
Gzip Compression: 2GB file. Seconds. Less is better. VMware Player won!
FFmpeg: Convert H.264 HD to NTSC DV. Seconds. Less is better. VMware Player won!
OpenSSL: RSA 4096-bit performance. Signs per seconds. More is better. VirtualBox won!
RAMspeed (Integer): MB/s. More is better. VMware Player won!
RAMspeed (Floating Point): MB/s. More is better. VMware Player won!
According to the test results, it appears that VMware Player has a slight edge over VirtualBox in terms of CPU and memory. As far as disk performance is concerned, it is not conclusive which one is better.
It may still be hard to generalize the findings reported here, since this is a case of one particular hardware setting with a limited set of tests. So take it with a grain of salt.
If you want to share your experience with VMware Player or VirtualBox in terms of their performance, feel free to share it in the comment section.