Trailing-Edge
-
PDP-10 Archives
-
bb-k345a-sb
-
lodtst.mem
There are 7 other files named lodtst.mem in the archive. Click here to see a list.
SIMPKG
Simulation Package for System Testing
Date: February 1978
File: SPST.RNO
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 and copied only in accordance with the terms of such
license.
Digital Equipment Corporation assumes no responsibility for the use
and reliability of its software on equipment that is not supplied by
Digital.
Copyright c 1978 Digital Equipment Corporation, Maynard, Mass.
Page 2
CONTENTS
Page
CHAPTER 1 INTRODUCTION 1-1
CHAPTER 2 DESCRIPTION OF PROGRAMS 2-1
2.1 LODTST 2-1
2.2 SIMULATED USER JOBS 2-1
CHAPTER 3 HOW TO RUN THE SIMULATION PACKAGE 3-1
3.1 MODULO OPTION 3-1
3.2 CATEGORY OPTION 3-2
3.3 COMMANDS DURING TEST OPERATION 3-3
3.4 END-OF-TEST PROCEDURE 3-3
CHAPTER 4 EXPLANATION OF ERROR MESSAGES 4-1
4.1 ERRORS FROM LODTST 4-1
4.2 ERRORS FROM SIMULATED USER JOBS 4-3
CHAPTER 5 INSTALLATION-DEPENDENT PARAMETERS 5-1
5.1 LODTST PARAMETERS 5-1
5.2 SIMULATED USER JOB PARAMETERS 5-2
CHAPTER 6 SAMPLE RUNS 6-1
6.1 SAMPLE RUN 1 6-1
6.2 SAMPLE RUN 2 6-2
CHAPTER 1
INTRODUCTION
This document describes a job simulation package developed by the
Academic Computer Center at Western Michigan University. The package
consists of a master control program and a number of user jobs that
simulate various types of user activity.
The objectives of the simulation package are:
1. To provide a known, reproducible system load that can be used
to evaluate system performance when software or hardware
changes are made.
2. To provide a very heavy, complex job mix that can be used to
verify system software reliability when a new monitor system
is loaded.
3. To provide a simulated system load that can be used to
isolate intermittent hardware failures that occur under the
TOPS-10 monitor but do not occur under field service
diagnostics.
The master control program (LODTST) reads user-specified test
parameters and performs all necessary details of test execution. It
logs in the desired number of simulated user jobs of each type, starts
the jobs executing, and prints SYSTATs at the specified interval.
For many tests, the SYSTATs provide sufficient information to evaluate
system performance. In other cases, additional information must be
recorded and printed using the DECsystem-10 System Performance
Evaluation Package (PERF).
A complete simulation run normally involves running SCDSET to define
scheduling parameters, LODTST to log in the simulated jobs, and LINES
to record performance data. To simplify operator interaction, the
test can be run under an OPSER.ATO file. The commands to run SCDSET,
LODTST, and LINES are stored in the .ATO file. Parameter data for
SCDSET and LODTST are read from disk files. Chapter 6 contains
examples of this technique.
CHAPTER 2
DESCRIPTION OF PROGRAMS
2.1 LODTST
The LODTST program is the master control program for the simulation
package. It reads the user-specified test parameters, logs in the
specified number of jobs of each type, and takes SYSTATs at the
user-specified interval.
LODTST logs in and RUNS all of the simulated user jobs using one
pseudo-TTY. The user jobs hibernate until LODTST has logged in all of
them. LODTST then wakes all of the jobs at the same time.
LODTST requests SYSTATs through a separate job that is controlled with
a pseudo-TTY. Both LODTST and its SYSTAT job run in a high-priority
queue to provide precise timing of SYSTATs.
At the end of the simulation run, LODTST prints an ending SYSTAT and
executes an end-of-test procedure that allows the simulated user jobs
to exit gracefully and delete any disk files they may have created.
For any problems encountered, LODTST logs an error message on its
controlling TTY. For severe errors, the entire simulation is
terminated. For less serious errors, the simulation continues.
Chapter 4 contains a complete list of all error messages from the
simulation package.
Section 2.2 describes the simulated user jobs supplied with the
package. Additional job types may be added at your installation's
discretion. Chapter 5 contains a description of the procedures for
defining job names and the areas they are to be run from.
2.2 SIMULATED USER JOBS
The simulated user jobs are made up of calls to standard routines
defined in a universal file. The standard routines select job size,
execute CPU loops, read and write disk records, open device channels,
and so forth. The ranges of job size, length of CPU loop, number of
disk records, and so forth, are determined by drawing random numbers
from uniform or exponential distributions and applying a base and
range factor defined by adaptation parameters contained in each user
program. Default values for each parameter are provided with the
simulation package. Chapter 5 contains instructions on how to modify
these parameters.
For any problems encountered by a simulated user job, an error message
is logged on the terminal associated with device OPR. For some error
conditions, the job will sleep for a defined interval and try the
function again later. For other error conditions, the job will
DESCRIPTION OF PROGRAMS Page 2-2
abandon the function for this cycle. In either case, the simulation
run will be allowed to continue. Chapter 4 contains a description of
the error codes printed by simulated user jobs and an analysis of
their causes and impact on test integrity.
A description of the program loop executed by each of the simulated
user jobs is given below. The end of this chapter contains a detailed
explanation of the various standard routines.
LODIO - I/O bound job.
1. Do core size selection (RANCOR)
2. Do disk I/O (RANDSK)
3. Repeat from step 1
LODCPU - CPU bound job.
1. Do core size selection (RANCOR)
2. Do uniform CPU loop (RANCPU)
3. Repeat from step 1
LODAVE - Average job.
1. Do core size selection (RANCOR)
2. Do uniform sleep (RANSLP)
3. Do uniform CPU loop (RANCPU)
4. Do disk I/O (RANDSK)
5. Repeat from step 1
LODPQ1 - Conversational job.
1. Do core size selection (RANCOR)
2. Do exponential sleep loop (EXPSLP)
3. Do exponential CPU loop (EXPCPU)
4. Repeat from step 1
LODSLP - Sleep job.
1. Do core size selection (RANCOR)
2. Do uniform sleep (RANSLP)
3. Repeat from step 1
LODCOR - Change high- and low-segment size.
1. Do high- and low-segment size selection (RANHGH)
2. Repeat from step 1
DESCRIPTION OF PROGRAMS Page 2-3
LODOPN - Open device channels.
1. Open a device (RANOPN)
2. Repeat from step 1
The following is a description of the standard routines in the LODUNV
universal file. They are called from the simulated user jobs supplied
with the package and may be used to create additional jobs at your
discretion.
RANCOR - Do core size selection.
Count the number of times entered and change the
low-segment core size every tenth time it is entered.
The formula for selecting new core size is:
core size = (RAN0 * CORRNG) + CORBAS + MINSIZ
where core size is in pages. RAN0 is a random number
in the range 0 to 1 from the uniform distribution.
CORRNG is the core size range parameter defined in the
calling program. CORBAS is the base core size.
MINSIZ is the size of the calling program.
The default values of 0 for CORBAS and 35.0 for CORRNG
imply a range of 0 to 35 pages plus MINSIZ.
RANDSK - Do disk I/O.
This routine writes a random number of records on the
public disk structure, reads the records back, and
compares the data to check for hardware failures. The
data written is a rotating bit pattern. The units
selected include all structures in the search list for
the area from which the program is run.
The number of words written is determined by the
formula:
words = (RAN1 * BLKRNG) + BLKBAS
where RAN1 is a random number in the range 0 to 1 from
a uniform distribution. BLKRNG is the
installation-dependent range parameter for disk
writes. BLKBAS is the base number of words.
The default values of 4096 for BLKRNG and 0 for BLKBAS
imply that a range of 0 to 4095 words are written (0
to 32 blocks).
The number of buffers used is determined by the same
random number as the number of words written.
buffer = (10 * RAN1) + 1
which implies a range of 1 to 10 buffers.
DESCRIPTION OF PROGRAMS Page 2-4
RANSLP - Do uniform sleep.
This routine executes a sleep monitor call with the
length of sleep determined by the formula:
sleep = (RAN2 * SLPRNG) + SLPBAS
where RAN2 is a random number in the range 0 to 1 from
a uniform distribution. SLPRNG is the
installation-dependent sleep range. SLPBAS is the
base sleep time.
The default values of 60.0 for SLPRNG and 0 for SLPBAS
imply a range of 0 to 59 seconds.
RANCPU - Do uniform CPU loop.
This routine executes a CPU loop of length:
length = ((RAN3 * CPURNG) + CPUBAS)/FACTOR
where RAN3 is a random number in the range 0 to 1 from
a uniform distribution. CPURNG is the
installation-dependent range factor. CPUBAS is the
installation-dependent base CPU length and FACTOR is a
scaling parameter to determine the number of SOJGs
that can be executed per second on a KA, KI, or KL.
FACTOR is determined automatically by the program,
depending on the processor type. With the default
values of 1 second for CPURNG and 0.5 second for
CPUBAS, this implies a range of 0.5 to 1.5 seconds.
EXPSLP - Do exponential sleep.
This routine executes a sleep monitor call with the
length of sleep:
sleep = MIN ((EXRAN2 * SLPRNX), SLPMAX)
where EXRAN2 is a random number in the range 0 to
infinity from an exponential probability density
function with a mean of 1. SLPRNX is the
installation-dependent desired mean sleep time.
SLPMAX is the maximum sleep time.
The default values of 6 seconds for SLPRNX and 60
seconds for SLPMAX imply a range of 0 to 60 seconds
with a mean of about 6 seconds.
EXPCPU - Do exponential CPU loop.
This routine executes a CPU loop of length:
length = MIN ((EXRAN3 * CPURNX), CPUMAX)/FACTOR
where EXRAN3 is a random number in the range 0 to
infinity from an exponential probability density
function with a mean of 1. CPURNX is the
installation-dependent desired mean CPU time. CPUMAX
is the maximum CPU time.
The default values of 0.1 second for CPURNX and 1.0
second for CPUMAX imply a range of 0 to 1 second with
a mean of about 6 ticks.
DESCRIPTION OF PROGRAMS Page 2-5
RANHGH - Do high- and low-segment size selection.
This routine changes the high- and low-segment size
every time it is entered.
The new size for high segment is (in pages):
high segment = (RAN0 * HGHRNG) + HGHBAS
The new size for low segment is (in pages):
low segment = (RAN1 * LOWRNG) + LOWBAS + MINSIZ
where RAN0 and RAN1 are random numbers in the range 0
to 1 from a uniform distribution. HGHRNG is the range
factor for high segment. HGHBAS is the base
high-segment size. LOWRNG is the range factor for low
segment. LOWBAS is the base low-segment size. MINSIZ
is the program size of the calling program.
The default values imply a range of high-segment size
from 0 to 32 pages and a range of low-segment size
from 0 to 32 pages plus program size.
RANOPN - Open device channels.
This routine opens and closes a device selected at
random from an installation-dependent list of devices
(DEVLST). The device is selected by the formula:
index = RAN1 * DEVNUM
where RAN1 is a random number in the range 0 to 1 from
a uniform distribution. DEVNUM is the number of
devices.
Note that the designations of RAN0, RAN1, RAN2, and RAN3 imply the
existence of four separate uniform random number generators. This is
to allow independent sets of random numbers for core size, sleep
length, CPU interval, number of words written to disk, and so forth,
for those programs that do several functions. The uniform
distributions all select a random number the first time they are
called, and the complement of that number the second time they are
called. That is, a random selection of n on an odd-numbered call
implies a selection of 1-n on the next call.
This procedure minimizes the amount of distortion in accumulated
runtime and other resources due to the random number selection
process. For any two jobs of the same class, the amount of CPU loop
executed and the number of disk reads and writes performed would be
identical, if the scheduler treated them equally, except for the
effect of the last random number selected. Differences due to random
number selection are not cumulative. This procedure gives good
randomness to the test with most differences between jobs of the same
class being attributable to the scheduler.
The exponential distributions are generated by the following formula:
EXRAN = -LOG(1-RAN)
where RAN is one of the uniform random number generators described
above. This generates a negative exponential probability density
function with a mean of 1. The exponential distributions yield a wide
range of values with a large number of small values and a few very
large values.
CHAPTER 3
HOW TO RUN THE SIMULATION PACKAGE
These examples demonstrate how to run the LODTST program from a TTY.
Examples with input from disk and execution from an OPSER.ATO file are
given in Chapter 6.
3.1 MODULO OPTION
The MODULO option logs in jobs starting at a base project-programmer
number (PPN), and increasing modulo a specified number. This allows
the test designer to specify different scheduling classes for
different PPNs to test the class scheduler.
To log in all jobs under the same PPN, specify a modulo number of 1 (0
defaults to 1).
1. Login a 1,2 job.
2. RUN TST:LODTST ;Specify PPN where LODTST is stored
3. ENTER INPUT FILE SPEC (DEFAULT IS TTY)
<CR> ;<CR> for TTY input
;FILE SPEC for disk input
4. ENTER INTERVAL BETWEEN SYSTATS (MINS)
5 ;For SYSTATS every 5 minutes
;<CR> for no SYSTATS
5. ENTER LENGTH OF TEST (MINS)
30 ;For 30-minute test
;<CR> for no time limit
6. ENTER JOB DATA OPTION (CATEGORY OR MODULO)
M ;M for MODULO
7. ENTER BASE PPN (PROJ,PROG)
4,220 ;Base area to LOGIN user jobs
8. ENTER MODULO NUMBER (DECIMAL)
3 ;For PPNs 4,220; 4,221; 4,222
;0 or 1 logs all jobs under 4,220
HOW TO RUN THE SIMULATION PACKAGE Page 3-2
9. LODIO: 3 ;3 jobs, 1 each under 4,220; 4,221; 4,222
LODCPU: 6 ;6 jobs, 2 each under 4,220; 4,221; 4,222
LODAVE: 9 ;9 jobs, 3 each under 4,220; 4,221; 4,222
LODPQ1: 3 ;3 jobs, 1 each under 4,220; 4,221; 4,222
LODSLP: 1 ;one job under 4,220
LODCOR: 2 ;2 jobs, 1 each under 4,221; 4,222
LODOPN: 3 ;3 jobs, 1 each under 4,220; 4,221; 4,222
10. [HH:MM:SS] SIMULATION STARTED. ;LODTST prints simulation
start time
11. Test is now up and running. For a list of commands that may
be typed, see Section 3.3.
Note that the MODULO counter is not reset between job categories.
Therefore, LODCOR started at 4,221 because one job was already counted
for LODSLP.
3.2 CATEGORY OPTION
The category option allows the test designer to log in any number of
jobs of each category under any number of specific PPNs.
Steps 1 through 5 are the same as steps 1 through 5 above.
6. ENTER JOB DATA OPTION (CATEGORY OR MODULO)
C ;C for category
7. ENTER PPN AND NUMBER OF JOBS (PROJ,PROG:NUMBER) ONE PER LINE,
EXTRA CARR RET TO GO TO NEXT CATEGORY.
LODIO: 4,220:1 ;1 job under 4,220
LODIO: 4,221:2 ;2 jobs under 4,221
LODIO: <CR> ;Go to next category
LODCPU: 4,221:3 ;3 jobs under 4,221
LODCPU: <CR> ;Go to next category
LODAVE: 4,223:4 ;4 jobs under 4,223
LODAVE: <CR> ;Go to next category
LODPQ1: 4,221:2 ;2 jobs under 4,221
LODPQ1: <CR> ;Go to next category
LODSLP: 4,220:1 ;1 job under 4,220
LODSLP: <CR> ;Go to next category
LODCOR: <CR> ;<CR> for no LODCOR jobs
LODOPN: 4,221:2 ;2 jobs under 4,221
LODOPN: <CR> ;Go to next category
8. [HH:MM:SS] SIMULATION STARTED. ;LODTST prints the simulation
start time
9. Test is now up and running. For a list of commands that may
be typed, see Section 3.3.
HOW TO RUN THE SIMULATION PACKAGE Page 3-3
3.3 COMMANDS DURING TEST OPERATION
While LODTST is running the simulation, it wakes at the specified
interval to request SYSTATs. When a SYSTAT is printed, LODTST types
an asterisk on its controlling terminal and rings the bell. LODTST
also monitors its controlling terminal for operator commands. The
commands accepted are:
SYSTAT ;Print a SYSTAT
EXIT ;Terminate the simulation and stop all jobs
KILL ;Terminate the simulation and log out all jobs
/EXIT ;Change end-of-test flag to EXIT option
/KILL ;Change end-of-test flag to KILL option
SYSTATs requested by the operator are in addition to any being
generated at the requested SYSTAT interval.
The EXIT command is used when the operator wants to stop the test and
does not care if the jobs are logged out. The exit procedure stops
the jobs at a point where they have no disk files open. This leaves a
clean disk system.
The KILL command performs all the functions of the EXIT command. In
addition, it logs out all simulated jobs. This will cause the
resources used by the simulated user jobs to be recorded by the
accounting system. Another simulation can be started without
reloading the monitor.
3.4 END-OF-TEST PROCEDURE
At the time specified by the length-of-test parameter, LODTST will
execute an end-of-test procedure. The procedure is either the KILL or
EXIT command described above, depending on the setting of end-of-test
flag as defined in the installation-dependent parameter (KILL).
The operator can dynamically change the setting of the end-of-test
flag with the /KILL or /EXIT commands.
The normal exit message is:
[HH:MM:SS] END OF EXIT PROCEDURE. ALL JOBS STOPPED. LODTST
EXITING.
CHAPTER 4
EXPLANATION OF ERROR MESSAGES
4.1 ERRORS FROM LODTST
All errors from LODTST are logged on its controlling TTY. Standard
format recovery implies reentering the parameter if the input is from
a TTY, and termination of the simulation if the input is from disk.
The standard line terminators are CRLF or altmode (ESC).
OPNERR - ? FAILURE OPENING INPUT DEVICE.
The specified input device does not exist. The
simulation run is terminated.
LOKERR - ? FAILURE ON LOOKUP FOR INPUT FILE.
The specified input file does not exist. The
simulation run is terminated.
OPTERR - ? ILLEGAL FORMAT, TYPE CAT OR MOD.
The specified job data option must be category or
modulo. Do standard format recovery.
CATERR - ? ILLEGAL CATEGORY SPECIFICATION.
TYPE PROJ,PROG:NUMBER (PROJ,PROG IN OCTAL, NUMBER IN
DECIMAL). PROJ is a project number, PROG is a
programmer number. PROJ and PROG must contain only the
digits 0 through 7 and be separated by a comma. After
PROJ,PROG there must be a colon followed by the number
of jobs to be logged in under the specified PPN. Do
standard format recovery.
PPNERR - ? ILLEGAL PROJ,PROG NUMBER FORMAT.
In the modulo option, the base PPN must contain only
the digits 0 through 7, and the two octal numbers must
be separated by a comma. Do standard format recovery.
MODERR - ? ILLEGAL MODULO NUMBER.
The modulo number must contain only the digits 0
through 9. Do standard format recovery.
EXPLANATION OF ERROR MESSAGES Page 4-2
CHKERR - ? ILLEGAL COMMAND. COMMANDS ARE:
SYSTAT, EXIT, KILL, /EXIT, /KILL.
An unrecognizable command was typed. Retype the
command. Type SYSTAT to get a SYSTAT on the line
printer. Type EXIT to stop all jobs, delete all files,
and exit. Type KILL to log out all jobs, delete all
files, and exit. Type /EXIT to set the end-of-test
flag to EXIT. Type /KILL to set the end-of-test flag
to KILL.
LOGERR - ? TIME LIMIT EXCEEDED LOGGING IN JOB UNDER PPN
NNNNN,NNNNN.
LODTST has issued a LOGIN request over the controlling
pseudo-TTY, but the subjob has not returned status bits
of job-number-assigned job-logged-in, TTY-at-monitor
level, and ready-for-a-command within the specified
time limit (approximately 60 seconds).
Possible problems are that this PPN is not allowed to
log in (not defined in ACCT.SYS), is in a class with no
quota, and so forth.
LODTST assumes the job will complete LOGIN and proceeds
to issue the RUN command. Should this assumption prove
to be false, the test will run without this job.
RUNERR - ? TIME LIMIT EXCEEDED TRYING TO RUN JOB LODXXX UNDER
PPN NNNNN,NNNNN.
LODTST has issued a RUN command over the pseudo-TTY,
but the subjob has not returned the status bits of
job-number-assigned, job-logged-in, at-user-level,
not-doing-TTY-output within the required interval (60
seconds).
Possible problems are that the job did not get logged
in properly (per LOGERR), or job type LODXXX does not
exist. LODTST assumes that the job started properly
and releases the pseudo-TTY. Should this assumption
prove to be false, the test will run without this job.
STSERR - ? JOBSTS UUO FAILED.
The JOBSTS monitor call used to determine if jobs are
logged in and running has failed. This should never
happen.
The simulation run is terminated.
GETSPR - ? ILLEGAL FILE SPEC.
The required format is:
DEVICE:FILENAME.EXTENSION[PROJ,PROG]
The simulation run is terminated.
EXPLANATION OF ERROR MESSAGES Page 4-3
PTYERR - ? OPEN FAILURE ON PTY.
The OPEN monitor call failed for the pseudo-TTY
channel.
Possible problems are that the monitor is not assembled
with pseudo-TTYs or none are available.
The simulation run is terminated.
FINERR - ? [HH:MM:SS] EXIT TIME LIMIT REACHED.
SOME LODTST JOBS MAY STILL BE RUNNING.
The end-of-test procedure failed to stop all jobs (EXIT
option) or some jobs did not LOGOUT within the time
limit (KILL option). The last job to KJOB sometimes
gets hung up in CONFIRM dialogue due to a race
condition in KJOB.
LODTST exits to monitor.
4.2 ERRORS FROM SIMULATED USER JOBS
The error messages from the simulated user jobs are sent to the
terminal associated with device OPR. The messages are of the format:
? [HH:MM:SS] JOB NNN "ERROR MESSAGE"
where NNN is the job number of the job having the error and "ERROR
MESSAGE" is one of the messages described below. The current time is
printed in square brackets.
When an error occurs, the job name is also changed to contain the
error number. For example, if a LODAVE job got the error indicated by
EO4, its name would be changed to EO4AVE.
Some errors cause the job to be delayed for a specified interval
before repeating the function that failed. The length of the delay is
an installation-dependent parameter (ERRSLP).
Note that any delayed jobs distort the distribution of CPU time, disk
reads and writes, and so forth, in the test results.
EO1 - CORE UUO FAILED.
Explanation - The amount of core requested exceeded CORMAX.
Either the installation-dependent core range
parameters are too large or the monitor value is
too low.
Recovery
Action - Delay and try again with different random number.
EO2 - DISK INIT FAILURE.
Explanation - Cannot open disk file to do disk I/O. Possible
problem getting core blocks from monitor core
block pool.
Recovery
Action Delay and try again.
EXPLANATION OF ERROR MESSAGES Page 4-4
EO3 - DISK ENTER FAILURE.
Explanation - Cannot write using this filename. Possibly file
already exists and is being written by some other
program under the same PPN, or cannot get core
blocks from monitor core block pool. (Note
LODTST system ensures unique filenames by making
job number part of filename.)
Recovery
Action Delay and try again.
EO4 - DISK OUTPUT FAILURE.
Explanation - Public structure full, cannot get monitor core
blocks, or possibly device hardware failure.
Recovery
Action - Delete file and exit from disk I/O subroutine.
Program will cycle to next function and continue
operation.
EO5 - DISK CLOSE FAILURE.
Explanation - Public structure full, cannot get monitor core
blocks, or possibly hardware failure.
Recovery
Action - Delete file and exit from I/O subroutine.
Program will continue operation.
EO6 - DISK LOOKUP FAILURE.
Explanation - Cannot look up file just written. Possibly
cannot get core blocks, or there are hardware
problems.
Recovery
Action - Program will cycle to the next function and
continue operation.
EO7 - DISK WRITE AND READ COMPARE FAILURE.
Explanation - Data read from disk is different from what was
written. Probably hardware failure.
Recovery
Action - Delete file and exit from disk I/O subroutine.
Program continues to operate.
EO8 - DISK INPUT FAILURE.
Explanation - Possible hardware error on device.
Recovery
Action - Delete file and exit from disk I/O subroutine.
Program continues to operate.
EO9 - DISK RENAME FAILURE.
Explanation - Could not delete disk file. Possible hardware
failure.
Recovery
Action - Exit from disk I/O subroutine. Program continues
to operate.
CHAPTER 5
INSTALLATION-DEPENDENT PARAMETERS
5.1 LODTST PARAMETERS
AREA - Contains the PPN where the simulated user jobs are
stored. Default value is [5,23] (TST:).
PRGNAM - A table containing the names of the simulated user
jobs. To add a new job type to the system, store an
EXE version of the program under the PPN defined in
AREA and enter the job's name as a new entry in this
table. The parameter NUMCLS defines the new length of
the table automatically.
KILL - This is an assembly parameter that is stored as the
initial value of the end-of-test flag. The default
value is zero, which gives the EXIT option. A nonzero
value gives the KILL option.
DEBUG - This is an assembly parameter that defaults to zero. A
nonzero value will include code to print any error
messages received from the subjob on the pseudo-TTY for
debugging purposes.
SPOOL - This is an assembly parameter that defaults to zero. A
nonzero value causes the SYSTATs to be spooled to disk
by issuing a SET SPOOL LPT command to the job doing the
SYSTATs. A zero value causes the SYSTATs to be printed
immediately on the line printer.
SYSPRJ - This is an assembly parameter that defines the project
number of the SYSTAT job logged in by LODTST. The
default value is 1.
SYSPRG - This is an assembly parameter that defines the
programmer number of the SYSTAT job. The default value
is 2. Note that both this PPN and the PPN of LODTST
must have high-priority queue privileges to ensure that
SYSTATs will be printed at accurate intervals.
INSTALLATION-DEPENDENT PARAMETERS Page 5-2
5.2 SIMULATED USER JOB PARAMETERS
There are a number of parameters associated with each of the simulated
user jobs. Although the same parameters exist in several jobs (for
example, in LODAVE, LODIO, LODCPU), they are defined separately in
each job. This is done so that an installation can generate any
number of job types by changing the parameter ranges and defining new
job names in PRGNAM. The meaning of the parameters is explained
below.
ERRSLP - Contains a decimal number defining the number of
seconds to sleep after an error has occurred before
trying the function again. This is the delay time
indicated in the description of error messages in
Chapter 4. The default value is 10 seconds.
Used in LODAVE, LODIO, LODCPU, LODSLP, LODPQ1, and
LODCOR.
CORBAS - Contains a floating-point number defining the base core
size in pages for the routine RANCOR. The default
value is 0.
Used in LODAVE, LODCPU, LODIO, LODSLP and LODPQ1.
CORRNG - Contains a floating-point number defining the core
range for the routine RANCOR. The default value is
35.0 pages.
Used in LODAVE, LODCPU, LODIO, LODSLP, and LODPQ1.
CORCNT - Contains a decimal number defining how often the
routine RANCOR actually changes core size. With the
default value of 10, RANCOR selects a new random core
size every tenth time.
Used in LODAVE, LODCPU, LODIO, LODSLP and LODPQ1.
HGHBAS - Contains a floating-point number defining the base
high-segment size for the RANHGH routine. The default
value is 0.0.
Used in LODCOR only.
HGHRNG - Contains a floating-point number defining the range of
high-segment core sizes for the routine RANHGH. The
default value is 32.0.
Used in LODCOR only.
LOWBAS - Contains a floating-point number defining the base
low-segment size for routine RANHGH. The default value
is 0.0.
Used in LODCOR only.
LOWRNG - Contains a floating-point number defining the range of
low-segment core sizes for the routine RANHGH. The
default value is 32.0.
Used in LODCOR only.
INSTALLATION-DEPENDENT PARAMETERS Page 5-3
MINSIZ - This parameter is set automatically by each program in
the initialization routine. It contains the initial
program size used by the RANCOR and RANHGH routines.
Current program sizes are shown below.
Program MINSIZ
LODIO 2
LODCPU 2
LODAVE 2
LODPQ1 2
LODSLP 2
LODCOR 2
LODOPN 2
BLKBAS - Contains a floating-point number defining the base
number of disk words to be written in routine RANDSK.
The default value is 0.0 word.
Used in LODIO and LODAVE.
BLKRNG - Contains a floating-point number defining the range of
words written in disk I/O routine RANDSK. The default
value is 4096.0 words.
Used in LODIO and LODAVE.
SLPBAS - Contains a floating-point number defining the base
sleep time for the routine RANSLP. The default value
is 0.0 second.
Used in LODAVE and LODSLP.
SLPRNG - Contains a floating-point number defining the range of
sleep times for the routine RANSLP. The default value
is 60.0 seconds.
Used in LODAVE and LODSLP.
CPUBAS - Contains a floating-point number defining the base CPU
time loop. The default value is 0.5 second.
Used in LODAVE and LODCPU.
CPURNG - Contains a floating-point number defining the range of
the CPU time loop. The default value is 1.0 second.
Used in LODAVE and LODCPU.
SLPRNX - Contains a floating-point number defining the desired
mean sleep time for the exponential sleep routine
EXPSLP. The default value is 6.0 seconds.
Used in LODPQ1 only.
INSTALLATION-DEPENDENT PARAMETERS Page 5-4
SLPMAX - Contains a floating-point number defining the maximum
sleep time for the exponential sleep routine EXPSLP.
The default value is 60.0 seconds.
Used in LODPQ1 only.
CPURNX - Contains a floating-point number defining the desired
mean CPU time for the exponential CPU routine EXPCPU.
The default value is 0.1 second.
Used in LODPQ1 only.
CPUMAX - Contains a floating-point number defining the maximum
CPU time loop for the exponential CPU routine EXPCPU.
The default value is 1.0 second.
Used in LODPQ1 only.
DEVLST - Contains the list of logical device names used by the
RANOPN routine. New entries may be added by inserting
the 6-bit device names into the table. The parameter
DEVNUM defines the new length of the table
automatically.
XITMES - Contains the KJOB command used in the KILL option of
the exit procedure. It is defined in LODUNV. It must
be of the form:
KJOB DSK:DUMMYX.LOG=/switches
The default switches are /K/Z:0/VD:D
CHAPTER 6
SAMPLE RUNS
6.1 SAMPLE RUN 1
This sample shows how to run LODTST with input from a disk file. The
file contains the responses to all LODTST dialogue questions. As the
responses are read from the file, they are echoed on the controlling
terminal along with the normal prompting messages.
The disk file used in the sample is named WILD.INP. It contains a set
of LODTST parameters that generate a very complex job mix that can be
used to verify system software reliability whenever a new monitor is
loaded. The file contents are as follows:
10 ;Interval between SYSTATs
60 ;Length of test
M ;Job data option
4,220 ;Base PPN
2 ;MODULO number
20 ;20 LODIO jobs
10 ;10 LODCPU jobs
10 ;10 LODAVE jobs
10 ;10 LODPQ1 jobs
4 ;4 LODSLP jobs
8 ;8 LODCOR jobs
2 ;2 LODOPN jobs
The dialogue to start this test is as follows:
1. Login a [1,2] job.
2. RUN TST:LODTST
3. ENTER INPUT FILE SPEC (DEFAULT IS TTY:)
WILD.INP ;Enter the name of the input file
All further responses are read from the disk file WILD.INP. The test
proceeds automatically from this point.
SAMPLE RUNS Page 6-2
6.2 SAMPLE RUN 2
This sample shows how to run a simulation test using an OPSER.ATO file
with SCDSET and LODTST reading their parameters from disk files. In
the sample, the operator copies the SCDSET input file TEST.SCD to the
standard input file SCDSET.INP. This technique makes it possible to
do multiple runs with the same job mix using different scheduling
parameters. For each run, the operator copies the desired SCDSET file
to the standard input file.
The test procedure is:
1. LOGIN a [1,2] job.
2. .COPY SCDSET.INP=TEST.SCD
3. .R OPSER
:AUTO TEST.ATO
The simulation test proceeds automatically.
The OPSER.ATO file TEST.ATO contains the following commands:
:SLOGIN
:DEFINE LI=
LI-SET HPQ 1
LI-R SCDSET
LI-INPUT
LI-GET SYS: LINES
LI-CST
LI-DET
:SLOGIN
:DEFINE M-
M-RUN TST:LODTST
M-TEST.INP
The LODTST input file TEST.INP contains the following:
10 ;Interval between SYSTATs
30 ;Length of test
M ;MODULO option
4,204 ;Base PPN
1 ;All jobs under 4,204
12 ;12 LODIO jobs
12 ;12 LODCPU jobs
28 ;28 LODAVE jobs
8 ;8 LODPQ1 jobs
0 ;No LODSLP jobs
0 ;No LODCOR jobs
0 ;No LODOPN jobs