Browse Prior Art Database

Software Monitor Data Reduction Program

IP.com Disclosure Number: IPCOM000079313D
Original Publication Date: 1973-Jun-01
Included in the Prior Art Database: 2005-Feb-26
Document File: 5 page(s) / 64K

Publishing Venue

IBM

Related People

Anderson, HA: AUTHOR [+2]

Abstract

In the operation of software system monitors, a block of combined task- and system- status information is written onto some form of secondary storage, usually tape, each time that control passes certain points in the supervisor. Examples of such points are when a terminal input is received, when a line is written to a terminal, when a system command is initiated, etc. This status information, however, constitutes a voluminous data base and is not suitable for subsequent data analysis. Accordingly, a transformation termed "data reduction" has to be carried out. Such transformation achieves the following: a) It eliminates unnecessary information such as, for example, text which is written to a terminal.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 30% of the total text.

Page 1 of 5

Software Monitor Data Reduction Program

In the operation of software system monitors, a block of combined task- and system- status information is written onto some form of secondary storage, usually tape, each time that control passes certain points in the supervisor. Examples of such points are when a terminal input is received, when a line is written to a terminal, when a system command is initiated, etc. This status information, however, constitutes a voluminous data base and is not suitable for subsequent data analysis. Accordingly, a transformation termed "data reduction" has to be carried out. Such transformation achieves the following:
a) It eliminates unnecessary information such as, for example,

text which is written to a terminal.
b) It groups records (each recorded event is a record) according

to their semantic meaning and accumulates data on the

characteristics of such groups. There are several levels of

details for such grouping, such as for example, user-session

level (from log-on to log-off, for example), user-transaction

level (from carriage return to system prompt,

for example), command level, etc.

It has been found that the writing of the data reduction program, as referred to hereinabove, is a nontrivial task and has to overcome the following difficulties:
a) Patterns corresponding to events on the various levels of

detail have to be recognized.
b) Since the action of a complex computer system is unpredictable,

the data reduction practice has to, therefore, be prepared to

accept any possible sequence of records and recover from

unexpected sequences as rapidly as possible (error handling).
c) New definitions of patterns of events and new transformations

on the data preferably have to be readily added.

Heretofore, the capability of software monitors has been substantially limited by the lack of a data reduction program of desirable generality. The problem which is presented is the transformation of the sequence of records collected by a software system (event tracing) monitor into suitable form for data analysis. To solve this problem, there is presented herein, a syntax driven algorithm for performing such transformation, the algorithm carrying out the transformation on several levels of detail. This algorithm encompasses the following features:
a) The use of a finite state automaton to recognize patterns.
b) A modular design with each module performing a well-defined

local transformation of the data (locality principle).
c) The use of the PL/I preprocessor for program generation and

program editing.

The salient idea of this algorithm is the considering of the string of monitor records as a sentence of a formal language, and the breaking up of such

1

Page 2 of 5

sentence into its constituents phrases by a suitable syntax driven parsing algorithm. In this connection, it has been found that a regular language is powerful enough and, therefore, a finite state automaton is a suitable algorithm for performing the desi...