Browse Prior Art Database

Flexible and Scalable Mechanism for Navigating Hierarchically Structured Measurement Data

IP.com Disclosure Number: IPCOM000114942D
Original Publication Date: 1995-Feb-01
Included in the Prior Art Database: 2005-Mar-30
Document File: 4 page(s) / 220K

Publishing Venue

IBM

Related People

Berry, RF: AUTHOR [+2]

Abstract

Many systems (e.g., operating systems) are organized into a part-subpart hierarchy. As a result, measurement data taken from such systems typically have a hierarchical structure that can provide useful insights when navigating the data. Existing approaches to automating the navigation of measurement data either lack flexibility in that only limited portions of the measurement hierarchy can be traversed or they lack scalability due to the large amounts of memory required to store inter-measurement links. Herein is disclosed a mechanism for navigation that constructs inter-measurement links dynamically so as to provide both flexibility and scalability.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 26% of the total text.

Flexible and Scalable Mechanism for Navigating Hierarchically Structured
Measurement Data

      Many systems (e.g., operating systems) are organized into a
part-subpart hierarchy.  As a result, measurement data taken from
such systems typically have a hierarchical structure that can provide
useful insights when navigating the data.  Existing approaches to
automating the navigation of measurement data either lack flexibility
in that only limited portions of the measurement hierarchy can be
traversed or they lack scalability due to the large amounts of memory
required to store inter-measurement links.  Herein is disclosed a
mechanism for navigation that constructs inter-measurement links
dynamically so as to provide both flexibility and scalability.

      To illustrate the problem addressed with this technique, a
running example is used throughout.  In this example, measurements
data are taken from multiple computer systems using a facility such
as RMF Monitor III.  Each measurement value is described in terms of
a metric (e.g., overall delay, memory delay, queue length), a
computer system on which the measurement is taken, and a workload
(e.g., an address space).  To simplify our later discussion, a very
modest amount of data is considered in which there are thirty-one
metrics, thirty-one systems, and thirty-one workloads.  The first
metric (denoted by index 0) is the most general (e.g., overall
delay), and the remaining thirty metrics are its components (e.g.,
processor delay, memory delay).  Similarly, the first workload
represents an aggregation of the other thirty workloads, such as a
domain with thirty address spaces; and the first system represents an
aggregation of the individual systems present (the sysplex).  A
measurement name is expressed as an ordered triple (M sub i, S
sub j, W sub k), where M denotes metric, S denotes system, W
denotes workload, and i, j, and k are zero-based indexes.
Note that there are (31) sup 3 approx 30,000 measurement names.

      Inter-measurement relationships induced by the part-subpart
hierarchy can be represented by a Measurement Navigation Graph (MNG).
Fig. 1 depicts portions of a MNG for the data in our example.  The
root of the graph specifies an indicator for the overall performance
of the thirty individual computer systems: (M sub 0, S sub 0,
W sub 0).  As shown in the Figure, (M sub 0, S sub 0, W sub 0) can be
broken into parts by metric, which is depicted by the arrows from
(M sub 0, S sub 0, W sub 0) to (M sub 1, S sub 0, W sub 0), ellipsis,
(M sub 30, S sub 0, W sub 0).  The latter can be partitioned as well;
for example (M sub 1, S sub 0, W sub 0) has the sub-components (M sub
1,
S sub 1, W sub 0), ellipsis, (M sub 1, S sub 30, W sub 0).

      The time and difficulty involved with interpreting measurement
data has motivated the development of applications that assist with
automating this process.  Typically, these applications employ a MNG
(at least implicitly...