Created a simple monitoring tool named wspy for workload spy. Source is hosted at http://www.github.com/mvermeulen/wspy

This is a wrapper program that collects data as the program runs. The initial program has two data collectors:

• trace – which uses the ftrace linux kernel tracer to turn on tracing for scheduler events for fork/exec/exit. It parses these to create and display a process tree.
• timer – which fires off a timer once per second to run a collector for /proc/stat contents

This program is somewhat different from the “trup” utility I created while at AMD. That program instrumentation used the “ptrace” capability to trace all processes in a tree and then instrument them. This approach might have the following advantages and disadvantages:

• Using capabilities like ftrace is hopefully less intrusive and hence less performance overhead than ptracing every process.
• ftrace seems to require a root setup; so the program runs as root, though I’ve added a “-u” option to set the user for the child.
• Instrumentation is based on the entire system; rather than individual processes. The pro/con of this is less precision for an individual application (particularly a single process) but a bigger picture of the overall system state.

wspy code is an initial code base that I can mutate and change for other tracing.

Following is an example of initial usage:

mev@popayan:~/wspy$ ./wspy -?
./wspy: fatal error: usage: ./wspy [CcFf][-r name][-u uid] ...
	-C	turn on CPU usage tracing (default = on)
	-c	turn off CPU usage tracing (default = on)
	-F	turn on kernel scheduler tracing (default = on)
	-f	turn off kernel scheduler tracing (default = on)
	-r	filter for name of process tree root
	-u	run  as user 
mev@popayan:~/wspy$ 

The following command line was run on the Phoronix Test Suite to run the c-ray application:

mev@popayan:~/wspy$ sudo ./wspy -o wspy.output.txt -r ./c-ray -u mev phoronix-test-suite batch-run c-ray

Following is a link to the output file created. The “-r” helps trim auxiliary process tree information.

Clearly more to do both in data output/presentation and maturity of this program. However, also useful to have a code base that runs on x86/ARM to now tweak for other instrumentation.