Minutes of 3/17/24 meeting

[Vara Prasad <prasadav at us dot ibm dot com>]

Minutes of 3/17/24 meeting

I left the meeting early hence i might not have captured all the 
discussions.

As usual please correct me or add details that i have missed.

-------------------------------------------------------------------------
Attendees: Will, Jim, Larry, Hien, Vara, Sri, Brad, Ananth, Elena

We started the discussion to see if the goal of systemTAP
is to help solve performance problems or debugging or both.
Everyone agreed that it should help in both the situations.

Should we allow both read and write. Write is useful
for debugging.
Jim mentioned that fault injection is a good application for debugging
purpose that probably needs write access and we should allow.
After some discussion we concluded that we should support
both read and write capability. We should support write
mode only for privileged users.
Brad: has volunteered to write-up the safety document
Vara: brought up the issue of transport mechanism. The issue
here is debugging might need high throughput and high
volume transport compared to Performance. Will quickly
pointed out that those who wants feed performance data into
external performance tools will have the same needs.
The consensus was very clear that transport needs
to handle high throughput and high volume.
Vara brought up the issue of tracing which is very useful
in debugging.
Vara brought up the issue if we should
support one level of calls in the tracing or multilevel.
Frank said we need to call graph analysis with static
code. Only problem might be we might have to put probes
in areas of the code that is not easy to reach. Sridhar
has agreed to drive this issue by doing a write up in
the mailing list.