Since a while OpenJDK has been available for Linux on the mainframe as an open source Java. In the last week I had two different performance situations with this. The reason is very simple: OpenJDK for z, doesn’t have a JIT. So it can be used e.g. for writing an installer but for everything that’s repeating the IBM JDK is the only choice.
The effect varies by workload, but it’s usually more that one order of magnitude. In the example above showing relative performance for an average workload the IBM Java does better even without the JIT and with the JIT it’s more than a factor of 50! I had to break up the axis to show it in one chart.