enira.net

We are all living in a technological wasteland.

RSS
people

Benchmarking VMware ESXi versus Citrix XenServer

Introduction

In this little post I’ll be comparing VMware ESXi and Citrix XenServer.

The base image consists from a Windows 10: 32-bit image. On this image I installed Passmark software CPU Bench. Each VM will have 2048MB RAM available and only one instance will be running on the hypervisor hosts.

The base score for the i3-5010u is Score: 3054, Single threaded score: 1178

Testing

For each test I will gradually up the CPU power. The P symbol means Passmark score, S is the equivalent of the Single Thread score.

Test 1: 1 vCPU

Test one will be a 1 virtual CPU, running a test on each machine.

XenServer: 1 vCPU	P: 910 S: 983	P: 905 S:975	P: 918 S:903	
VMware: 1 vCPU		P: 957 S: 1146	P: 964 S:1151  	P: 951 S:1153

You can see that in the single threaded score there is a penalty of around 15-18% for Citrix Xen Server. VMware does a better job with only around 3% loss.
As for the Passmark test, we can see that VMware does a better job than XenServer. However the result is quite small. VMware performs arounf 5% better on a single core VM.

Test 2: 2 vCPUs

2 virtual cores with one core per socket (2×1)

XenServer: 2 vCPU	P: 1583 S:993 	P: 1633 S:991 	P: 1643 S:985
VMware: 2 vCPU		P: 1826 S:1150	P: 1826 S:1155 	P: 1829 S: 1155

In this test we can see that VMware seems to keep a rock steady score for the single threaded performance. Citrix XenServer stays the same too. Adding more virtual CPUs doesn’t seem to add any extra advantage to the single threaded performance.
However we can see in this test that with adding an extra CPU to both virtual machines, the passmark score seems to double. Good evolution, we are getting there.

Test 3: 4 vCPUs

This test is a little bit odd. XenServer allows a setup of 4 virtual cores on one CPU. And these results are in the same range as the 2×1 test. As the i3-5010U has 2 cores and 2 threads on each. This proves that XenServer will offer more options than VMware, though it’s worth checking your hardware layout before throwing around cores.

4 virtual cores with one core per socket

XenServer: (4x1) 4 vCPU	P: 1691 S: 987	P: 1668 S: 988 	P: 1732 S: 995
XenServer: (2x2) 4 vCPU	P: 2180 S: 987  P: 2110 S: 991  P: 2131 S: 993
VMware: (2x2) 4 vCPU	P: 2337 S: 1157 P: 2335 S: 1161 P: 2339 S: 1162

This test is the final score. Noticable, the single thread score doesn’t improve much. You can see the Citrix Xen Server stays the lowest of the two.

All results

single

passmark

Conclusion

You can see Citrix XenServer stays below VMware in raw performance. Is Citrix XenServer useless? No, despite the lower CPU performance, each hypervisor does more than the ones found in VMware. In VMware you will need a vCenter running to perform some of the advanced features. Example: In Citrix XenServer environment the configuration concerning high availability is found in each hypervisor.
This could explain some of the differences found in performance.

Leave a Reply

You must be logged in to post a comment.