Browse Prior Art Database

High Performance System Component Event Tracing

IP.com Disclosure Number: IPCOM000104861D
Original Publication Date: 1993-Jun-01
Included in the Prior Art Database: 2005-Mar-19
Document File: 4 page(s) / 95K

Publishing Venue

IBM

Related People

Rankin, TF: AUTHOR [+3]

Abstract

A high performance mechanism for tracing operating system components is disclosed. Intended to work in conjunction with the MVS/ESA* Servicability Level Indication Processing Facility (SLIP), this tracing mechanism establishes a single entry point that would be branched to by any component module when the criteria for the event trace is met. This scheme would enable SLIP trap tracing for all events within a component by providing a single place to set a SLIP trap.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

High Performance System Component Event Tracing

      A high performance mechanism for tracing operating system
components is disclosed.  Intended to work in conjunction with the
MVS/ESA* Servicability Level Indication Processing Facility (SLIP),
this tracing mechanism establishes a single entry point that would be
branched to by any component module when the criteria for the event
trace is met.  This scheme would enable SLIP trap tracing for all
events within a component by providing a single place to set a SLIP
trap.

      For serviceability, operating system components need a means of
tracing the sequence of "events" that are occurring on the system.
An event, as used in this context, is defined to be any sequence of
instructions that is considered significant to the component.  It
could be as small as one instruction (e.g., an IPTE instruction) or
it could be a large set of instructions (e.g., obtaining a system
control block).  In general, an end user's request for some system
function would be fulfilled by the execution of many events.

      In MVS, event tracing is accomplished using the Component Trace
Facility.  A trace scenario is invoked using an operator command,
defining what events are to be traced.  Hooks in the component
determine when a particular event is being traced and what data is to
be collected for the event.  When the criteria for the event is met,
the data is collected in a trace buffer to be dumped, examined, and
debugged later.

      The disadvantage of this technique is that the increased path
length is often prohibitive to performance-sensitive components, such
as...