Trailing-Edge
-
PDP-10 Archives
-
scratch
-
10,7/unscsp/lodtst/perfor.mem
There are 7 other files named perfor.mem in the archive. Click here to see a list.
PERF
System Performance Evaluation Package
Date: February 1978
File: SPAP10
Version: 1
The information in this document is subject to change without notice
and should not be construed as a commitment by Digital Equipment
Corporation. Digital Equipment Corporation assumes no responsibility
for any errors that may appear in this document.
The software described in this document is furnished under a license
and may be used or copied only in accordance with the terms of such
license.
Digital Equipment Corporation assumes no responsibility for the use or
reliability of its software on equipment that is not supplied by
DIGITAL.
Copyright C 1978 by Digital Equipment Corporation
Page 2
CONTENTS
Page
PREFACE
CHAPTER 1 INTRODUCTION 1-1
CHAPTER 2 MONITOR DATA COLLECTION CODE AND PROGRAM
DESCRIPTIONS 2-1
2.1 PART 1--COLLECTING PERFORMANCE DATA(MONITOR
DATA COLLECTION CODE) 2-1
2.2 PART 2--RECORDING DATA ON DISK (LINES.MAC) 2-1
2.3 PART 3--PRODUCING GRAPH, REPORTS, AND
SUMMARIES 2-2
2.3.1 SIMP1.FOR 2-2
2.3.2 GRAPH1.FOR 2-2
2.3.3 LNSRPT.FOR 2-3
CHAPTER 3 PROGRAM OPERATION 3-1
3.1 HOW TO START RECORDING PERFORMANCE EVALUATION
DATA 3-1
3.2 HOW TO EXAMINE THE LINES JOB FOR PERCENTAGE OF
CPU TIME USED 3-1
3.3 HOW TO PRINT THE RESOURCE UTILIZATION GRAPH 3-1
3.4 HOW TO PRINT THE PERFORMANCE EVALUATION
REPORTS 3-2
CHAPTER 4 REPORT DESCRIPTIONS 4-1
4.1 RESOURCE UTILIZATION GRAPH 4-1
4.2 RESOURCE UTILIZATION REPORT (RPTA) 4-2
4.3 PQ1 REQUEUE DATA (RPTB) 4-3
4.4 PROCESSOR QUEUE DATA (RPTC) 4-3
4.5 SPECIAL QUEUE DATA (RPTD) 4-4
4.6 CLASS PERCENTAGES (RPTE) 4-5
4.7 SUMMARY REPORT (RPTS) 4-5
APPENDIX A EXAMPLE GRAPH, REPORTS, AND SUMMARIES A-1
FIGURES
FIGURE A-1 Resource Utilization Graph A-3
A-2 Resource Utilization Report (RPTA) A-4
A-3 PQ1 Requeue Data (RPTB) A-5
A-4 Processor Queue Data (RPTC) A-6
A-5 Special Queue Data (RPTD) A-7
A-6 Class Percentages (RPTE) A-8
A-7 Summary Report (RPTS) A-9
Page 3
CONTENTS (CONT.)
Page
TABLES
TABLE 2-1 Installation-Dependent Parameters 2-3
4-1 Resources Recorded on the Resource Utilization
Graph 4-1
A-1 Primary and Secondary Percentage Allocations
of Scheduler Classes A-2
Page 4
PREFACE
This document describes the DECsystem-10 System Performance Evaluation
Package. This package was developed by the Western Michigan
University (WMU) Academic Computer Center.
The performance evaluation package consists of three parts:
1. Monitor code that allows you to collect performance data.
2. A user-mode program that allows you to record the data on
disk.
3. A number of user-mode programs that allow you to produce a
graph, reports, and summaries.
The objectives of the performance evaluation package are:
1. To provide information on the usage of important system
resources.
2. To provide a useful tool for analyzing system performance.
Chapter 1 is an introduction. Chapter 2 describes all of the programs
used in the performance evaluation package and the monitor data
collection code. Chapter 3 tells you how to operate the programs.
Chapter 4 describes the graph, the reports, and the summaries that
make up Part 3 of the performance evaluation package. The Appendix
contains examples of the graph, the reports, and the summaries.
CHAPTER 1
INTRODUCTION
This document describes the DECsystem-10 System Performance Evaluation
Package. This package was developed by the Western Michigan
University (WMU) Academic Computer Center.
The performance evaluation package consists of three parts:
Part 1--Monitor code that allows you to collect performance
data. (See Section 2.1.)
Part 2--A user-mode program that allows you to record the data
on disk. (See Section 2.2.)
Part 3--A number of user-mode programs that allow you to
produce reports, summaries, and a graph. (See Section 2.3.)
The objectives of the performance evaluation package are:
1. To provide information on the usage of important system
resources.
2. To provide a useful tool for analyzing system performance.
When you use the performance evaluation package, you can collect and
record data at the following levels:
Level 1--contains the data required to produce the Resource
Utilization Graph (Figure A-1) and the Resource Utilization
Report, RPTA (Figure A-2).
Level 2--contains all of the data in level 1 plus the
additional data required to produce the reports that allow you
to tune the scheduler (that is, PQ1 Requeue Data, RPTB (Figure
A-3), Processor Queue Data, RPTC (Figure A-4), Special Queue
Data, RPTD (Figure A-5), and Class Percentages, RPTE (Figure
A-6)).
You may print the above reports individually or in groups if the
monitor is assembled at level 2.
In addition to the reports listed above, you may produce a Summary
Report, RPTS (Figure A-7), to summarize the data in RPTA, RPTB, RPTC,
RPTD, and RPTE at any interval.
Chapter 4 contains a complete description of each of these reports.
CHAPTER 2
MONITOR DATA COLLECTION CODE AND PROGRAM DESCRIPTIONS
This chapter describes Parts 1 through 3 of the performance evaluation
package.
Section 2.1, which describes the monitor code that allows you to
collect performance data, corresponds to Part 1. Section 2.2, which
describes the user-mode program that allows you to record the data on
disk, corresponds to Part 2. Section 2.3, which describes a number of
user-mode programs that allow you to produce a graph, reports, and
summaries, corresponds to Part 3.
2.1 PART 1--COLLECTING PERFORMANCE DATA(MONITOR DATA COLLECTION CODE)
The standard monitor tables contain the data for level-1 recording.
Therefore, level-1 recording does not require any new code or monitor
tables. To perform level-1 collection, assemble the monitor by
setting the FTPSCD assembly switch to 0.
The SCDPER monitor table contains the data for level-2 recording.
GETTAB 142 (GTSPA under feature test switch FTPSCD) reads this data.
To perform level-2 collection, assemble the monitor by setting the
assembly switch FTPSCD to -1.
2.2 PART 2--RECORDING DATA ON DISK (LINES.MAC)
The LINES program starts as a detached job running in [1,2] (or any
number with SPY privileges). Then, once per minute, LINES records the
monitor performance data. Every 5 to 10 minutes, LINES adds the
accumulated data to the end of a disk file. The name of the disk file
is in the form
mmmdd.DAT
where
mmm is a 3-character alphabetic representing the month
dd is a 2-character numeric representing the day of the month
For example, on 9 March the filename is MAR09.DAT.
MONITOR DATA COLLECTION CODE AND PROGRAM DESCRIPTIONS Page 2-2
You can cause LINES to assemble in such a way that it can record
either level-1 or level-2 data on the disk if the monitor is assembled
at a sufficient recording level. To do this, set the FTSCHED assembly
switch to the following values:
1. To record level-1 data, set FTSCHED to 0.
2. To record level-2 data, set FTSCHED to -1.
LINES changes its own program name once per minute to display the
percentage of the central processor (CPU) time used in the last minute
(this is for reading from SYSTAT, and so forth).
2.3 PART 3--PRODUCING GRAPH, REPORTS, AND SUMMARIES
Sections 2.3.1 through 2.3.3 describe the user-mode programs that
allow you to produce a graph, reports, and detailed summaries.
2.3.1 SIMP1.FOR
The SIMP1 program is the preprocessor for the GRAPH1 program. It
compresses LINES data into a number of contiguous 20-minute intervals
with date headings.
2.3.2 GRAPH1.FOR
The GRAPH1 program produces an 8-1/2- by 11-inch graph of 24 hours of
data (7:00 am to 7:00 am). This data consists of the following:
1. The amount of CPU time used.
2. The number of lines used.
3. The rate of swapping.
4. The rate of reading and writing on the public disk structure.
To tailor the scaling of the Resource Utilization Graph to your
installation, you must supply various installation-dependent
parameters and then reassemble the GRAPH1 program. GRAPH1 contains
the following two sets of parameters:
Set 1 - The parameters that determine the maximum scale values
used in plotting.
Set 2 - The parameters that determine the scale values used in
printing the maximum percentage for any 1-minute
interval.
Table 2-1 lists the Set 1 and Set 2 installation-dependent parameters.
MONITOR DATA COLLECTION CODE AND PROGRAM DESCRIPTIONS Page 2-3
Table 2-1
Installation-Dependent Parameters
Parameter Description Default Value
Set 1 Set 2
ABSMAX(1) MXLT Maximum number of 64 lines
active terminal lines
ABSMAX(5) MXST Maximum swapping rate 36,000 blocks/minute
ABSMAX(4) MXDT Maximum disk I/O rate 5,000 blocks/minute
2.3.3 LNSRPT.FOR
The LNSRPT program formats the data from LINES to provide detailed
reports. There are six detailed reports. They are:
1. Resource Utilization Report (RPTA)
2. PQ1 Requeue Data (RPTB)
3. Processor Queue Data (RPTC)
4. Special Queue Data (RPTD)
5. Class Percentages (RPTE)
6. Summary Report (RPTS)
CHAPTER 3
PROGRAM OPERATION
This chapter contains step-by-step instructions that tell you how to
operate the program. Each section tells you how to perform a
different task. These instructions are in the form of a sample
dialogue between you and the system. In every case, anything the
system prints is underscored.
3.1 HOW TO START RECORDING PERFORMANCE EVALUATION DATA
.GET SYS: LINES
.CSTART
.DETACH
3.2 HOW TO EXAMINE THE LINES JOB FOR PERCENTAGE OF CPU TIME USED
(Displayed in Job Name for LINES))
.SYS[1,2]
9 DET 51% 3+SPY SL SW 59
3.3 HOW TO PRINT THE RESOURCE UTILIZATION GRAPH
.RUN SIMP1
*JAN,25,1977 ;graph(s) will be printed for 24-hour periods
;from 7:00 am this day
*JAN,25 ;to 7:00 am this day
*EXIT
.RUN GRAPH
DATE IS JAN,25
.PRINT GRAPH.DAT
PROGRAM OPERATION Page 3-2
3.4 HOW TO PRINT THE PERFORMANCE EVALUATION REPORTS
.RUN LNSRPT
DAY(MMMDD): JAN25 ;data filename for day desired, in the form
;mmmdd
FROM(HR,MIN): 12,01 ;HR, MIN of beginning of period desired
TO(HR,MIN): 12,15 ;HR, MIN of end of period desired
;defaults for "FROM" and "TO" are 0,0 and
;24,00
OPTION: RPTA ;options are RPTA, RPTB, RPTC, RPTD, RPTE,
RPTS
OPTION: RPTB ;ALL for all the above, HELP for help
OPTION: RPTS ;RPTS for summary option
OPTION: <CR> ;end with a blank line
SUMMARY EVERY: 10 ;summary period for RPTS (default: 60
;minutes)
OUTPUT FILES: SCHEDA.DAT SCHEDB.DAT SCHEDS.DAT
.PRINT SCHED?.DAT
CHAPTER 4
REPORT DESCRIPTIONS
This chapter explains the format and content of the various reports.
For an example of each of the reports, see Figures A-1 through A-7 in
the Appendix.
4.1 RESOURCE UTILIZATION GRAPH
There are four resources recorded on the Resource Utilization Graph
(Figure A-1). Each of these resources is identified by a unique
letter code. (See Table 4-1.)
Table 4-1
Resources Recorded on the Resource Utilization Graph
Code Description Default Value
C Amount of CPU time used
L Number of terminal lines used 64 lines
S Swapping rate 36,000 blocks/minute
D Disk I/O rate 5,000 blocks/minute
In addition, the code letter O appears wherever two or more resources
overlap.
The vertical scale on the graph represents the percentage of resources
used. Figure A-1 is plotted using the maximum values indicated in
Table 4-1 as 100 percent. However, you may specify other values as
parameters in the graph program to change the scale. In other words,
your 100 percent need not be based on the same maximum values as the
100 percent in Figure A-1.
The maximum swapping rate of 72,000 blocks per minute is based solely
on the hardware transfer speed; it does not take into consideration
either hardware- or software-latency times. Therefore, you may not
specify more than 72,000 blocks per minute, no matter how many block
sequences are transferred.
REPORT DESCRIPTIONS Page 4-2
On the other hand, the disk rate of 5,000 blocks per minute is based
on single-transfer 128-word blocks. This figure is ultraconservative
if larger blocks are written (as occasionally happens).
The maximum of 64 lines is based on the total number of terminal ports
that are available to users.
The data is plotted for 20-minute intervals from data that is recorded
at 1-minute intervals. In other words, this is a summary of a
20-minute plot.
The horizontal scale on the graph represents the time of day. It
begins at 7:00 am of the indicated day and ends at 7:00 am of the
following day.
At the bottom of the graph several maximum values are printed to show
the highest values of the indicated variable for any recorded 1-minute
interval during the day. Typically, these values are larger than any
point plotted on the graph. This is because the graph plots a summary
(average) of a 20-minute interval. Where applicable, the user-defined
maximum value is indicated after the observed maximum percentage.
Remember, if you wish, these values can be different from the scale
values used in Figure A-1.
4.2 RESOURCE UTILIZATION REPORT (RPTA)
RPTA contains the minute-by-minute data from which the Resource
Utilization Graph is plotted, plus a number of additional data items.
RPTA includes the following data items:
Data Item Meaning
LINES The number of dial-up lines currently in use.
JOBS The number of job slots currently in use.
CPU-MIN(%) The percentage of the CPU time used during
the last minute.
OVRHEAD(%) The percentage of monitor overhead used
during the last minute.
DSK-MIN The number of 128-word blocks read and
written on the public file structure during
the last minute (expected maximum 5,000
blocks per minute).
SWP-MIN The number of 128-word blocks swapped to and
from memory during the last minute.
TTY-BUF(%) The percentage of buffer pool space used
during the last minute.
C.B.(%) The percentage of the memory blocks pool used
during the last minute.
SWAP-K(%) The percentage of swapping space used during
the last minute (includes dormant segments).
REPORT DESCRIPTIONS Page 4-3
4.3 PQ1 REQUEUE DATA (RPTB)
RPTB contains data on the requeues that usually go to PQ1 (unless the
job is in HPQ). RPTB includes the following data items:
Data Item Meaning
DTASL Number of times that a job was put to sleep
to wait for a DECtape to position during the
last minute.
MTASL Number of times that a job was put to sleep
to wait for a magnetic tape to position
during the last minute.
EWCNT Number of jobs requeued because an event wait
was satisfied.
TISJB Number of jobs requeued because a terminal
input wait was satisfied.
TOSJB Number of jobs requeued because a terminal
output wait was satisfied.
PISJB Number of jobs requeued because a
pseudo-terminal input wait was satisfied.
POSJB Number of jobs requeued because a
pseudo-terminal output wait was satisfied.
REQSS Number of jobs requeued because a SLEEP was
satisfied.
REQWK Number of jobs requeued because of a WAKE
monitor call (includes HIBERNATE for a number
of conditions).
REQJSD Number of jobs requeued because of a service
completion by DAEMON.
REQPQ1 Number of jobs requeued to PQ1 for all
reasons.
4.4 PROCESSOR QUEUE DATA (RPTC)
RPTC contains data on the activity in the various processor queues.
RPTC includes the following data items:
Data Item Meaning
XPQRP1 Number of jobs in PQ1 that expired their
quantum runtime during the last minute.
XPQRP2 Number of jobs in PQ2 that expired their
quantum runtime during the last minute.
XPQRHP Number of jobs in HPQs that expired their
quantum runtime during the last minute.
REPORT DESCRIPTIONS Page 4-4
Data Item Meaning
XPICP1 Number of jobs in PQ1 that expired their
in-core protect time during the last minute.
XPICP2 Number of jobs in PQ2 that expired their
in-core protect time during the last minute.
XPICHP Number of jobs in HPQs that expired their
in-core protect time during the last minute.
SWPKP1 Number of K swapped in for PQ1 jobs.
SWPKP2 Number of K swapped in for PQ2 jobs.
SWPJP1 Number of PQ1 jobs swapped in.
SWPJP2 Number of PQ2 jobs swapped in.
SWPJHP Number of HPQ jobs swapped in.
RNTPQ1 CPU seconds used by PQ1 jobs.
RNTPQ2 CPU seconds used by PQ2 jobs.
RNTHPQ CPU seconds used by HPQ jobs.
4.5 SPECIAL QUEUE DATA (RPTD)
RPTD contains data on the activity in the special queues. RPTD
includes the following data items:
Data Item Meaning
SWAP TIMES The time required to swap in a PQ1 or CMQ job
from the time it enters PQ1 to the time
swap-in is completed.
NUMBER The number of observations of swap time
recorded by the monitor during the last
minute.
MEAN The average swap time for the above
observations.
STDV The standard deviation for the above
swap-time observations.
WASTE CORE The amount (in pages) of memory occupied by
jobs that are not runnable (that is, in
long-term wait, terminal I/O wait, SLEEP,
STOP, and so forth).
NUMBER The number of observations of wasted core
recorded by the monitor during the last
minute.
REPORT DESCRIPTIONS Page 4-5
Data Item Meaning
MEAN The average waste core (in pages) for the
above observations.
STDV The standard deviation for the above
waste-core observations.
4.6 CLASS PERCENTAGES (RPTE)
RPTE contains data on the percentage of accumulated CPU time used by
the various classes. RPTE includes the following data item:
Data Item Meaning
CLAS N The percentage of accumulated CPU time used
by class N.
4.7 SUMMARY REPORT (RPTS)
The data contained in RPTA, RPTB, RPTC, RPTD, and RPTE is overwhelming
if printed minute by minute for a whole day. Therefore, RPTS provides
a convenient way to summarize the data in the reports at any interval.
(See Figure A-7.)
APPENDIX A
EXAMPLE GRAPH, REPORTS, AND SUMMARIES
This appendix contains an example of the Resource Utilization Graph
(Figure A-1) and examples of each of the following performance
evaluation reports:
1. Resource Utilization Report. (See Figure A-2.)
2. PQ1 Requeue Data, RPTA. (See Figure A-3.)
3. Processor Queue Data, RPTC. (See Figure A-4.)
4. Special Queue Data, RPTD. (See Figure A-5.)
5. Class Percentages, RPTE. (See Figure A-6.)
The data was collected on the Western Michigan University (WMU)
computer system during a 24-hour period. The Resource Utilization
Graph was plotted from data collected over the entire 24 hours. The
data contained in RPTA, RPTB, RPTC, RPTD, and RPTE was collected
during the 15-minute period from 12:01 pm to 12:15 pm. RPTS
summarized the data at 10-minute intervals.
The WMU computer system consisted of the following:
1. A KI10 with 160K memory.
2. Six RP02 disk packs and two RP03 disk packs on one channel.
3. Two RD10 swapping disks on one channel.
4. Two TU20 tape drives on the I/O bus.
5. A 6.03 monitor with the WMU class scheduler configured for 74
jobs and 16 scheduler classes.
Table A-1 lists the various classes along with their primary and
secondary percentage allocations.
EXAMPLE GRAPH, REPORTS, AND SUMMARIES Page A-2
Table A-1
Primary and Secondary Percentage Allocations of Scheduler Classes
Percentage Allocations
Class
Primary Secondary
0 95 95
1 3 3
2 1 1
3 1 1
EXAMPLE GRAPH, REPORTS, AND SUMMARIES Page A-3
Figure A-1 Resource Utilization Graph
EXAMPLE GRAPH, REPORTS, AND SUMMARIES Page A-4
Figure A-2 Resource Utilization Report (RPTA)
EXAMPLE GRAPH, REPORTS, AND SUMMARIES Page A-5
Figure A-3 PQ1 Requeue Data (RPTB)
EXAMPLE GRAPH, REPORTS, AND SUMMARIES Page A-6
Figure A-4 Processor Queue Data (RPTC)
EXAMPLE GRAPH, REPORTS, AND SUMMARIES Page A-7
Figure A-5 Special Queue Data (RPTD)
EXAMPLE GRAPH, REPORTS, AND SUMMARIES Page A-8
Figure A-6 Class Percentages (RPTE)
EXAMPLE GRAPH, REPORTS, AND SUMMARIES Page A-9
Figure A-7 Summary Report (RPTS)
EXAMPLE GRAPH, REPORTS, AND SUMMARIES Page A-10
Figure A-7 Summary Report (RPTS) (Cont.)
EXAMPLE GRAPH, REPORTS, AND SUMMARIES Page A-11
Figure A-7 Summary Report (RPTS) (Cont.)
EXAMPLE GRAPH, REPORTS, AND SUMMARIES Page A-12
Figure A-7 Summary Report (RPTS) (Cont.)