The contents of a thread's call stack at one point in time. The contents consist of a stack of frames, one for each method that was called.

CPU time is time spent running on a CPU. It does not include time spent waiting for I/O, sleeping, or anything else an application does between two events.

Short for stack frame. This refers to one node in a call trace. Each frame is associated to one method which was called.

Software component that collects performance data on a running application. It consists of a dynamic library that is loaded by the JVM. The JVM and the profiler agent communicate via the JVMPI interface. The JVM is instructed to load the profiler agent with the -Xrun command line option. See the JVMPI documentation for more information.

A frame that does not have a parent frame.

A mode of data collection where the profiler agent records the call stack of each thread in the application periodically. The resulting statistics include the number of times each call trace was sampled. Samples mode collects less information than times mode, but has a lesser amount of overhead.

The profiler agent may be invoked in suspended mode. When in suspended mode, it collects a minimum amount of data and produces a minimum of overhead until it is started through its JNI interface by calling com.fruitioninc.mjp.JniInterface.startProfiler().

A mode of data collection where the profiler agent times each and every method call. The resulting statistics include the accumulated CPU time and wall time of each call trace and the number of times the call trace occurred. Times mode collects more information than samples mode, but has a greater amount of overhead.

Wall time is total elapsed time. It includes time spent running on the CPU, waiting for I/O, sleeping, and anything else an application does between two events. The term "wall" is borrowed from somewhere now forgotten. It implies time as measured by a clock on the wall.