Looking at run time efficiencies on zOS 16 on 3931 mainframe with JAVA. We found on jobs running assembler level programs that specifying a large number of buffers when reading huge files reduces run times. Does the JVM exploit the zOS DFSMS subsystem directly?

We know that specifying BUFNO=50 for large files read by other languages improves run time by reducing the number of physical I/O operations. We are designing testing for JAVA batch jobs with these changes in JCL BUFNO values, but we do not know how well the JVM interacts with the zOS host.