For Linux VM, VMware tested the following three scenarios multiple times on three different large simulated customer network topologies:

  1. IP Availability Manager and the Global Manager running in the same VM.

  2. IP Availability Manager and the Global Manager running in individual VMs on the same ESX Server.

  3. IP Availability Manager and the Global Manager running in individual VMs on different ESX Servers.

    Using the midpoint results of scenario 1 as a baseline, Test results of scenarios 2 and 3 relative to scenario 1, expressed in percent presents the test results of scenarios 2 and 3 relative to scenario 1. “Min” represents the best results, “Max” represents the worst results, and “Median” represents the midpoint results.

Table 1. Test results of scenarios 2 and 3 relative to scenario 1, expressed in percent
Configuration Platform Min/Median/Max Discovery End-to-end

Elapsed

time

CPU

Elapsed

time

CPU

The VM running on the ESX Server

Red Hat Enterprise Linux AS/AP 5 (64- bit)

Min -4.1% -3.6% -2.9% -1.2%
Median 1.0% 0.6% 3.5% 2.2%
Max 9.2% 10.8% 8.1% 8.7%

Two VMs, one

running on one ESX Server, and the other running on another ESX Server

Red Hat Enterprise Linux AS/AP 5 (64- bit)

Min -16.1% -12.8% -16.5% -14.4%
Median -2.1% -2.3% 0.7% 1.7%
Max 2.8% 5.5% 3.2% 9.0%

In general, performance was negatively impacted when using two VMs on the same ESX Server. The performance showed a slight improvement when using two VMs on different ESX Servers.

When using two VMs on the same ESX Server, median end-to-end elapsed time increased 3.5% on Linux When using two VMs on different ESX Servers, with a no real difference (0.7% increase) on Linux.