2. PERFORMANCE REPORTING ANALYSIS › 2.5 Virtual Storage Management Analysis › 2.5.4 Virtual Storage Concepts in MVS/XA › 2.5.4.7 Allocation Measurements and Their Interpretation
2.5.4.7 Allocation Measurements and Their Interpretation
For portions of virtual storage like CSA, SQA, and the User
Region, a series of different measurements are available in
the type 78, subtype 2, Monitor I Activity record written by
RMF. This subtype record contains one Common Storage data
section in which are recorded systemwide measures, and
possibly one or more Private Area data sections if RMF has
been instructed to track the use of virtual storage by
particular jobs.
The following sections discuss the types of measurements
recorded by RMF virtual storage monitoring.
SIZE OF XXX
Size of xx refers to the size limit in KB for xxx, where xxx
represents CSA, SQA, User Region, or some other named portion
of virtual storage for which MVS sets a size limit. This
measurement represents the maximum size specified by the
installation in SYS1.PARMLIB, via exits, or in some other
fashion for this particular named portion of virtual storage.
ALLOCATED PAGES
Allocated Pages refers to the number of KB of storage
allocated by the Virtual Storage Manager for GETMAINs within
xxx, where xxx represents CSA, SQA, User Region, or some
other named portion of virtual storage.
These allocations occur in 4KB increments. One or more
GETMAINs may be satisfied by a single allocated block of
virtual storage.
For a GETMAIN to succeed, there must be a contiguous block of
virtual storage large enough to satisfy the GETMAIN. If
there is currently no such block allocated, the Virtual
Storage Manager (VSM) will attempt to find a free block of
virtual storage that is large enough, and will allocate this
block of storage.
If VSM is unable to find an allocated block with sufficient
space which has not yet been GETMAINed or a free block of
sufficient size, the function which depends upon this GETMAIN
may fail. If the portion of virtual storage required is in a
user address space, the user's job step may abnormally
terminate. If the portion of virtual storage required is
from a critical system area like CSA or SQA, then MVS may
invoke emergency swapping until this shortage is relieved.
FREE PAGES
Free Pages refers to the number of KB of storage not
currently allocated by the Virtual Storage Manager for
GETMAINs within xxx, where xxx represents CSA, SQA, User
Region, or some other named portion of virtual storage.
These unallocated portions of virtual storage occur in 4KB
increments. See the distinction described earlier for the
meaning of this measurement for SQA.
ALLOCATED AREA
The Allocated Area represents the number of KB between the
lowest allocated address of virtual storage and the highest
allocated address within xxx, where xxx represents CSA, SQA,
User Region, or some other named portion of virtual storage.
This measure will include all free pages that lie between
allocated pages.
This measurement will always be at least as large as the
measure Allocated Pages, since it includes all allocated
pages, plus any free pages scattered among those which are
allocated. The degree to which the measure Allocated Area
exceeds the measure Allocated Pages is an indication of
fragmentation of virtual storage.
LARGEST FREE BLOCK
The Largest Free Block refers to the number of KB of virtual
storage contained in the largest continuous block of storage
not currently allocated by the Virtual Storage Manager for
GETMAINs within xxx, where xxx represents CSA, SQA, User
Region, or some other named portion of virtual storage.
Since any GETMAIN requires a contiguous virtual address
range, this measure represents the size of the largest
GETMAIN that could succeed within xxx.
A simplified example of these measurements, using a
hypothetically very small CSA, is depicted in Figure 2-179,
which shows the following measurements:
o Size Of CSA 52 K bytes
o Allocated Pages of CSA 28 K bytes
o Free Pages of CSA 24 K bytes
o Allocated Area of CSA 44 K bytes
o Largest Free Block of CSA 12 K bytes
+-----------------------------+ === <--- 52K BYTES
|////////// Free //// 4K /////| |
+-----------------------------+ | === <--- 44K BYTES
| Allocated | | |
| Allocated 8K | | A
+-----------------------------+ | L ===
|////////// Free /////////////| S L |
|////////// Free /// 12K /////| I O LARGEST FREE BLOCK
|////////// Free /////////////| Z C | (12K BYTES)
+-----------------------------+ E A ===
| Allocated | T
| Allocated 12K | O E ALLOCATED FREE
| Allocated | F D PAGES PAGES
+-----------------------------+ (K BYTES) (K BYTES)
|////////// Free //// 4K /////| C A 8 4
+-----------------------------+ S R 12 12
| Allocated 4K | A E 4 4
+-----------------------------+ | A 4 4
| Allocated 4K | | | ---- ----
+-----------------------------+ | === 28K 24K
|////////// Free //// 4K /////| |
+-----------------------------+ ===
Figure 2-179. Allocation Measurements
MISCELLANEOUS MEASUREMENT CONSIDERATIONS
Most of the measurements described above are obtained via RMF
interval driven sampling as minimums, maximums, and totals.
Where minimums and maximums are maintained, there are
corresponding time stamps to indicate when the minimum or
maximum occurred.
RMF does not ordinarily sample virtual storage utilization as
frequently as it samples other quantities in the system. The
default RMF virtual storage sampling interval is set to be
ten times as long as the RMF sampling interval for other
measurements.
Totals can be used in conjunction with a data element
containing the number of samples to calculate averages for
the measurement period.
Since RMF samples its selected jobs for the Private Area data
sections only when that job is executing and swapped in,
various Private Area data sections may have different sample
counts. The field SMF78SAM, found in the RMF Product section
of the record, should be used when calculating averages from
totals in the Common Storage data section. In contrast, each
Private Area data section contains its own sample count,
R782SAMP, used which should be used when calculating averages
based upon totals in any particular Private Area data
section.
There is one other concern relating to sampling intervals.
Many of the values accumulated during the RMF sampling
process for virtual storage monitoring are accumulated by RMF
as single precision floating point numbers. The nature of
floating point numbers dictates that if a relatively small
number is added to a very large number, because of scale
differences between the two numbers, some or all of the
significance of the smaller number may be lost. This could
lead to understating various sampled values if the RMF
virtual storage sampling interval is set to too low a value.
Setting the sample interval to too low a value would cause
the accumulated totals to become larger as more frequent
samples are added to them, eventually leading to a
substantial difference in magnitude between the total being
accumulated and any given sample value being added to it. It
is partly for this reason that the default RMF virtual
storage sampling interval is set to be ten times as long as
the RMF sampling interval for other measurements. Of course,
reducing the RMF virtual storage sampling interval will also
increase the processor overhead due to sampling.
