IBM System/370 145 Manual page 93

tasks to reduce paging activity. Lat.er, when paging activity becomes
sufficiently low, the deactivated partition or region is reactivated.
CPU Speed. An improperly balanced relationship between CPU speed and
paging device speed can also caU~ie the system to become I/O-bound as a
result of paging. A Model 145 can execute a certain number of
instructions during the time re~lired to service a page-in request using
a given direct access device typ4~. A Model 145 can execute many more
instructions during a page-in from a 2314, for example, than can a Model
135. As long as there is usefu1 work for the CPU to perform while
paging operations occur, the sys1:em is not kept waiting for paging I/O.
However, if the concurrently ope]~ating programs are constantly executing
instructions faster than the
pag4~s
they require can be brought into real
storage, an excessively high paging rate could develop i.f task
deactivation were not invoked. ][n general, therefore, the larger scale
System/370 models require faster paging devices to handle a particular
page fault rate than do the smaller scale models.
Real Storage Size. The amount:. of real storage present in a system
affects the number of page fault~; that occur when a given job stream is
processed. If the amount of real storage present in the system is equal
to the total amount of virtual st:.orage being used by the concurrently
executing tasks, no page fau1ts ()ccur for programs that have been
fetched and initiated. When the amount of real storage. present is less
than the amount of virtual stora~Je being used, page faults occur. The
total number of page faults that occur for a given job stream is
affected by the ratio of virtual storage used to real storage available.
Assuming the amount of virtual. storage used in a given system remains
the same, t,he virtual-to-real storage ratio can vary. This occurs while
a given system experiences variat:.ions in the amount of real storage
actually available for paging as the amount of fixed real storage
changes during job stream process;ing. The real storage available for
paging at any point in time is the difference between the amount of real
storage in the system and the tot~al amount of long- and short-term fixed
real storage. For the IBM-supplied virtual storage operating systems,
the total amount of fixed real st~orage at any given time is the sum of
the:
• Resident (fixed) control
pro~~am
size, which does not vary after IPL
• Amount of long-term. fixed reall storage required for control blocks,
which can change as the level. of multiprogramming changes (OS/VSl
and OS/VS2 only)
• Amount of short-term fixed rE!al storage required for outstanding I/O
operations that have virtual channel programs, which fluctuates with
the I/O activity of the systE!m
• Amount of long-term fixed rea.l storage required by the job steps
executing in nonpaged (real) mode, if any
• Amount of long-term fixed real storage required by programs that
operate in paged mode but tha.t have a portion of their partition or
region always fixed (TCAM in OS/VS1 and OS/VS2, BTAM in DOS/VS, for
example).
As
the virtual-to-real storage ratio of a job stream increases, so
usually does the page fault rate. In general, the page fault rate
increases slowly for a while. At some point, the increase in page
faults begins .rising rapidly as the virtual-to-real storage ratio
continues to increase. Figure 15.15.2, shown later, illustrates the
general relationship between the number of page faults and the virtual-
to-real storage ratio.
A Guide to the IBM System/370 Model 145
83