Browse Prior Art Database

Real Storage/Paging Model

IP.com Disclosure Number: IPCOM000099919D
Original Publication Date: 1990-Mar-01
Included in the Prior Art Database: 2005-Mar-15
Document File: 7 page(s) / 291K

Publishing Venue

IBM

Related People

Jennings, JT: AUTHOR

Abstract

A method is disclosed for approximating a program's steady-state real storage usage for varying workload levels and varying degrees of system level storage contention. Given workload arrival rates, the program's virtual storage is divided into subsets, each consisting of pages that are referenced with the same frequency.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 20% of the total text.

Real Storage/Paging Model

       A method is disclosed for approximating a program's
steady-state real storage usage for varying workload levels and
varying degrees of system level storage contention. Given workload
arrival rates, the program's virtual storage is divided into subsets,
each consisting of pages that are referenced with the same frequency.

      This method is only applicable for steady-state analysis, i.e.,
items of work from the program's workloads arriving at steady rates.
Real storage activity for transient events, or bursts of activity, is
not as interesting, since the program will probably spend the
majority of its time in steady state.

      A least-frequently-used (LFU) page selection algorithm is
assumed.  In reality, however, a least-recently-used (LRC) algorithm
is probably used by the operating system.  In steady state, from the
program's perspective, the pages that it references least frequently
are the most likely candidates for page-out.  Therefore, the
assumption of an LFU page selection algorithm is adequate for steady
state.

      The program's workloads arrive at steady rates during steady
state.  In processing items of work from its workloads, the program
must reference pages from its virtual storage.  The set of pages that
the program references for an item of work may differ from one
workload to another.  To represent the program's behavior in
referencing virtual storage, the program's total set of virtual
storage pages is divided into subsets, each subset consisting of
pages that are referenced with the same frequency.

      Because the number of virtual storage pages required by the
program may depend on the operating environment (system
configuration, network configuration, etc.), it is useful to separate
the set of virtual storage pages into two groups: a set of base
pages, whose number is independent of the environment, and set of
data pages, whose number is dependent on the environment.  The method
for approximating real storage usage starts by calculating the
references per page per second for the base pages and data pages.
Once calculated, the reference rates are used to approximate real
storage usage with varying paging rates for the program's pages.

      A program's set of base pages consist of the pages which
contain the program code (or modules).  Miscellaneous data is
probably also contained in the base pages as well. The number of base
pages does not vary with the environment.

      In processing an item of work from a workload, the flow of
control of the program may pass through several modules and,
therefore, may reference several of the base pages. The different
workloads of the program probably use some common modules and some
unique modules.  To determine the module reference pattern for a
given set of workloads, it is useful to characterize the modules by
which workloads reference them.  Once this is done, it is possible to
partition the bas...