Previous Table of Contents Next


SUBJECT INDEX

2k experimental designs, 280, 283–292
2k–p experimental designs, 314–326
2kr experimental designs, 293–313
22r experimental designs, 293–308
6502 processor, xxvii, 166
68000 processor, xii, 275, 276, 281, 282, 359,
360, 361, 365, 367
Z8002 processor, 167, 360, 361, 365
8080 processor, 16, 276
8086 processor, xii, 275, 281, 282, 359
80286 processor, 282
Abnormal events monitoring, 97–98
Absolute deviation, mean, 193, 194, 197
Accounting logs, 114, 116–121, 176
analysis of, 117–119
data recorded in, 117–119
interpretation of, 117–119
use of, 119–121
workoad characterization using, 121
Accuracy vs. correctness, 32
Ackermann benchmark, 50, 53, 167, 358,
359, 360, 361, 365
SIMULA listing of, 53
ACM, 8, 9
Ada, 359
Addition instruction, 48
Additive model, 303–306
Advertising, by monitors, 105
AFIPS, 9
Agglomerative clustering techniques, 87
Aggregate subnetwork, 614, 615, 618
Airline reservation systems, 570
ALGOL–like languages, 53
ALGOL programs, 54
Algorithm(s):
convolution, 590, 593–607
correlation, 606–607
event–set, 408
page replacement, 382
paging, 404
Allocation of variation, 226–228, 295–298, 331–332, 372–374
Alternatives comparison:
several, 327
two, 208–209, 218
ALU (arithmetic logic units), 23, 61, 62, 66
Amdahl’s law validation, 257, 258
American Federation of Information
Processing Societies, see AFIPS
Analysis, 106
bottleneck, 563–568
cache, 404
complex, 19
coverage, 112
data, 97, 102
factor, 77
mean value, 570, 575–579
principal–component, 76–80
stress–load, 132
user participation in, 397
Analysis rat holes, 161
Analysis tools, commercial, 175
Analytical modeling, when to use, 31
Analyzer, in monitors, 106
ANOVA (analysis of variance), 250–253, 332–334, 374
ANOVA table:
one–factor designs, with equal sample
sizes, 333
for simple linear regression, 252
for two–factor design, without replication, 374
two–factor designs, with replication, 334
Antibugging, 416
Application benchmarks, 51–52
Applied Computer Research, 10
Approximate MVA, 579–585
Approximation:
for chi–square quantiles, 638
for F–quantiles, 638
for normal quantiles, 198, 638
for statistical tables, 638
Arc sine transformation, 261
ARGUS operating system, 230–232, 239, 240, 256
Arithmetic logic units, see ALU
Arithmetic mean, 70, 73, 188, 305
Arrival rate, effective, 512, 513–514, 536–537, 538, 539
Arrivals, 507–508, 510, 513–515, 551
bulk Poisson, 510
discouraged, 540
Poisson, 508
response–dependent, 621
train, 621
Artifact, 94
Art of:
data presentation, 139–163
performance evaluation, 7
workload selection, 60–69
Assertion checking, 112
Assumptions’ validation, 421
Asymmetric plot, 199, 200
Asymptotic bounds, 564–567
Autocovariance, 432, 465–466
Autoregressive sequence, 444
Availability, 34, 40, 43
Averaging, 73, 182
queue length, 429–430
utilization, 190
waiting time, 429–430
Balance, job flow, 551
Balanced job bounds, 585–591
Balanced system, 585
Bandwidth, 39
Banking systems, 56–57, 570
Bar chart, 140, 144
Baseline validation, 126
Base system for ratio games, 165–167
Batch computer systems, 594
Batches, 427–428
Batch means, 427–428, 431, 432
Batch monitors, 95
BCMP networks, 551
Benchmark(s), 47, 52 See also Workload
Ackermann, 50, 53, 167, 358, 359, 360, 361, 365
application, 51–52
debit–credit, 57, 175
Dhrystone, 55, 345, 346, 357
Doduc, 58
eqntott, 58
espresso, 58
fpppp, 59
GCC, 58
LINPACK, 55
LISP interpreter, 58
matrix, 58
NASA7, 58
puzzle, 50, 167, 359, 360, 361, 365
sieve, 50, 52, 175, 345, 346, 357, 359
SPEC, 58
spice, 58
tomcatv, 59
towers of Hanoi, 167, 360, 361, 365
TPC, 58
Whetstone, 54, 55
Benchmarking, 52
games, 130–131
mistakes in, 127–130
Bernoulli distribution, 483–484, 499, 500
use of, 483–484
variate generation, 480, 483–484
Bernoulli trials, 483
Beta distribution, 484–485
use of, 484
variate generation, 480, 485
Biased goals, 15
Bimodal distributions, 183
Binary n–cube networks, 433
Binomial distribution, 485–486, 499, 500
use of, 486, 492–493
variate generation, 480, 486
Birth–death processes, 516, 519–522
Bits, random, 457
Blocking, 552, 621
Bottleneck analysis, 563–568
Bottleneck device, 34, 558, 563
Bounds:
asymptotic, 564–567
balanced job, 585–591
Box–Cox transformations, 262–263, 264
Box–Mueller method, 494
bps (bits per second), 38
Broken scales in column charts, 149
Buckets, see Cells or buckets, in histograms
Buffers, 509
in monitors, 96, 128
in queues, 534–540
Bulk arrivals, 510
Poisson, 510
Bulk service, 510
Busy period, 524
Cache(s):
analysis, 404
comparison, 345–347
design alternatives, 352
disk, 574
hit ratios, 187
miss ratios, 187
in multiprocess environment, 353
Caching effects, 128
CACI, 502
Calendar queues, 410
Capacity, 39
knee, 39, 125
nominal, 39, 125
usable, 125
Capacity management, 123
Capacity planning, 123–127, 176
commercial products, 176
problems in, 125
steps in, 124
Categorical predictors in regression models, 254–257
Categorical variables, see Qualitative variables
Cauchy variate, 480
CDF, see Cumulative Distribution Function
Cells or buckets, in histograms, 75
size, 149, 197–198
Central limit theorem, 182
Central processing unit (CPU), 4
Central server model, 552
Central subsystem, 586
Central tendencies, indices of, 183
Certification, component, 132
Chandy–Herzog–Woo theorem, 615
Channels, 26
Characteristic polynomial, 445
Characterization method, 480–481
Chart(s):
Gantt, 149–152, 163
graphic, 140, 141–145, 176
bar, 140
column, 140, 145
broken scales in, 149
guidelines for preparing, 141
line, 140
misuse of, 145
mistakes in, 163
mistakes in preparing, 144–145
Schumacher, 160
Checklist:
for avoiding mistakes, in evaluation, 22
graphic charts, 143
for regression analysis, 269–270
for simulations, 398
Chi–square distance, 86
Chi–square distribution, 486–487
quantiles of, 632–633
approximation for, 638
use of, 486
variate generation, 480, 486–487
Chi–square test, 461–462
CISC (Complex Instruction Set Computer), 16
Classes of jobs, 548, 551
Clock, simulation, 503
Clock–advancing mechanisms, 407
Closed models, 400–401
Closed queueing networks, 547–548
Clustering analysis, 83–86, 175
agglomerative techniques, 87
divisive techniques, 87
interpretation, 90
problems with, 90–91, 175
CMG (Computer Measurement Group), 10, 176
Code:
optimization, 114, 120
restructuring, 120
Code size comparison study, 167, 329, 365, 390
Coefficient:
of correlation, 267
of determination, 227, 240, 246, 267
of multiple correlation, 246, 249
of skewness, 197
Coefficient of Variation (C.O.V.), 74, 181, 197
Collinearity, 253–254
Column chart, 140, 144, 145
broken scales in, 149
Commercial products:
for capacity planning, 176
for monitoring, 175
for performance analysis, 175
Comparators, 98
Comparison of systems, 203–220
several, 327
two, 211, 218, 327
Completeness, of metrics, 35
Complex analysis, 19
Component certification, 132
Components, principal, 77
Component under study (CUS), 61, 62
Composition method, 493. See also Decomposition method
Compression, data, 97
Computation of errors, 372
Computer network(s):
combining performance of layers, 187
comparison of algorithms, 218
comparison of architectures, 168
congestion control, 35, 414, 415
flow control, 592
gateway, 524, 556, 568
Kiviat graphs for, 159
local area, 39, 415, 621
metrics for, 35, 159
power, 592
queueing models for, 592
requirement specification, 42
simulation, 414–415
workload for, 63–65, 77, 82, 621
COMTEN, 175
Conclusion validation, 421
Concurrent simulation, 503
Conferences, performance, 8
Confidence coefficient, 204
Confidence interval(s), 147, 204, 219
for mean, 204–207
one–sided, 219
two–sided, 219
meaning of, 206
for model parameters, 229–232, 298–299, 335–337, 351–353, 374–379
for model predictions, 232–234, 299–301
one–sided, 213–215
for proportions, 215–216
one–sided, 219
two–sided, 219
for regression parameters, 267
vs. hypothesis testing, 213
Confidence level, 204
choosing a value for, 212
Configuration monitoring, 108
Configuration summarizing, 202
Confirmation experiment, 300
Confounding, 318–320
algebra of, 320–321
Congestion control:
metrics selection, 35
simulation model, 414, 415
Consistency tests, 420
Console, 102–103
functions, 109
Constant standard deviation of errors, 237
Contention, server, 621
Continuity test, 418–419
Continuous distributions, relationship among, 500
Continuous–event models, 399
Continuous simulation languages, 403
Continuous–state models, 399
Continuous–state processes, 515–516
Continuous–time models, 399
Continuous variables, 140
Contrasts, 285, 299, 301, 336
Convolution, 596
Convolution algorithm, 590, 593–607
Convolution method, 479–480
Correctness vs. accuracy, 32
Correlated observations, variance of the mean, 430–431
Correlated predictor variables, 268
Correlation, 181
coefficient of multiple, 246
Correlation coefficient, 181
Cost/performance ratio, 40
Counters, 97
overflow, 97
C.O.V., see Coefficient of Variation
Covariance, 181, 465
Coverage analysis, 112
CPM operating system, 282
CPU, see Central processing unit
CPU keelboat Kiviat graph, 155
CPU scheduling, 404
simulation, 398, 407, 423–424, 428–430, 433
Crashes, time between system, 214
Crossbar interconnection networks, 289–290
C/70 processor, 167, 360, 361, 365
CSMP, 403
Cumulative distribution function (CDF), 180.
See also Distributions; particular distributions
Curvilinear regression, 257
CUS (component under study), 61, 62
Customers, in a queue, 507, 510
Cycle, regeneration, 434
Cycle length, 438
Databases, Kiviat graphs for, 159
Database systems, 6, 12, 201, 242, 545
Data deletion, initial, 424–425
Datametrics, 176
Data presentation, 20, 25, 102, 106–108, 139–163
Data structures:
calendar queues, 410
heap, 410, 411
indexed linear list, 409
ordered linked list, 409
tree, 410, 411
dB (decibels), 262
Deadlock, 503
Deadlock prevention algorithms, 404
Debit–credit benchmark, 57, 175
pseudocode, 57
Decibels (dB), 262
Deciles, 194
Decision maker, 17
Decision makers’ games, 162
Decomposition, hierarchical, 608, 613–620
Decomposition method, 478–479
Defections from a queue, 622
Degeneracy tests, 419
Degrees of freedom, 194, 228, 250, 299, 300, 332, 348, 374, 489
Delay center, 509, 534, 551, 553, 563, 576, 586, 602
Deletion, initial data, 424–425
Demands, skewness of, 128
Dendrogram, 89
Density function, probability, 180
Design, resolution, 321–326
Designated subnetwork, 615
Destination, 35
Detail, level of, 18, 66–67, 394
Determination, coefficient of, 240
Deterministic distribution, 509
Deterministic models, 400
for verification, 417
Deviation, mean absolute, 193, 194, 197
Device, 510
bottleneck, 558, 563
Device demands, skewness of, 128
Device homogeneity, 552
Dhrystone benchmark, 55–56, 345, 346, 357
DIPS (Dhrystone instructions per second), 55
Discouraged arrivals, 540
Discrete distributions, relationship among, 500
Discrete event:
models, 399
simulation, 406–408
components of, 407–408
languages, 403
Discrete state:
models, 399
processes, 515–516
Discrete uniform distribution, 498
use of, 498
variate generation, 498
Discrete variables, 140
Disk cache, modeling effect of, 574
Dispersion, indices of, 193
selecting, 195–196
Displays, graphic, 418
Distance metrics, 86
chi–square, 86
Euclidean, 86
Distributed simulation, 503
Distributed–system monitors, 101–109
Distributed systems, 101–109, 200, 568
Distribution of data, determining, 196–199
Distribution function, cumulative, 180
Distributions:
asymmetric, 199, 200
commonly used, 483–501
Bernoulli, 483–484, 499, 500
use of, 483–484
variate generation, 480, 483–484
beta, 484–485
use of, 484
variate generation, 480, 485
binomial, 485–486, 499
use of, 492–493
variate generation, 480, 486
chi–square, 486–487
approximation for, 638
quantiles of, 632–633
use of, 486
variate generation, 480, 486–487
deterministic, 509
Erlang, 201, 487–488, 509
use of, 497
variate generation, 488
exponential, 488–489, 509, 510
truncated, 523–524, 531
use of, 488
variate generation, 475, 477, 489
extreme value, 477
F, 489–490
approximation for, 638
percentiles of, 636
use of, 490
variate generation, 490
gamma, 490–491
use of, 490
variate generation, 480, 490
general, 509
geometric, 200, 491
use of, 491
variate generation, 477, 491
hyperexponential, 509, 510
K–S (Kolmogorov–Smirnov), quantiles of, 637
Laplace, variate generation, 480
logistic, variate generation, 477
lognormal, 492
use of, 492
variate generation, 480, 492
negative binomial, 492–493
use of, 486, 492–493
variate distribution, 493
variate generation, 493
normal, 493–494, 499, 500
approximation for, 198–199, 638
area under, 628, 638
quantiles of, 629, 630
standard, 181–182, 493
test for, 237, 302
unit, 181–182, 493
use of, 494
variate generation, 480, 494
Pareto, 201, 495
use of, 495
variate generation, 477, 495
Pascal, 495–496
use of, 495
variate generation, 480, 496
Poisson, 201, 499, 500
use of, 486, 493, 496
variate generation, 480, 496
relationships among, 500
t, 497
quantiles of, 631
variate generation, 497
triangular, variate generation, 480
uniform (continuous), 497–498
use of, 497
variate generation, 498
uniform (discrete), 498
use of, 498
variate generation, 498
Weibull, 499
use of, 499
variate generation, 477, 499
Distributivity, k, 452, 467–470
Divisive clustering techniques, 87
Documentation, 397, 437
Doduc benchmark, 58
Domain of monitors, 94
Double–whammy graph, 147
Doubly–linked list, 409
Downtime, 40
Drivers, load, 132
Dynamic allocation of storage, 404
Dynamic memory management, 408
Dynamic models, 400
DYNAMO, 403
Editing sessions, 74, 122
EDP performance review, 10, 176
Educational environments, program usage in, 74
Effective arrival rate, 512, 513–514, 536–537, 538, 539
Effects:
computation of, 285–286, 294, 328–329, 369–372
confidence intervals for, 335–337, 351–353
Efficiency, programs, improvement of, 176
Efficiency of multiprocessor systems, 39
8080 processor, 166, 276
8086 processor, xii, 275, 281, 282
80286 processor, 282
Empirical distributions, generation, 475–476
Emulation, 403
keyboard monitor, 176
remote terminal, see Remote–terminal emulator
workstation, 138
Encryption, performance of, 243, 271
End systems, 35
Eqntott benchmark, 58
Erlang distribution, 201, 487–488, 509
use of, 487
variate generation, 480, 488
Erlang’s C formula, 530
Erlang’s loss formula, 536
Erroneous analysis, 18, See also Games;
Mistakes; Misuses
Error, 34, 222
standard, 205, 219
Error–free seconds, 40
Error monitoring, 108
Errors, computation of, 372
Errors in input, 19
Espresso benchmark, 58
Euclidean distance, 86
weighted, 86
Evaluation:
mistakes in, 14–22
performance, 8, 10, 16
avoiding mistakes in, 22
checklist, 22
systematic approach to, 22–26
technique(s), 16
selection of, 24, 30–33
Event, 94
notices, 408
routines, 407
scheduling, 407, 503
trace, 418
Event–driven monitors, 95
Event–driven simulation, 407
Events, independent, 179
Event–set algorithms, 408, 502
Exclusion, mutual, 621
Executable workload, 67
Execution profile, 112
Exerciser loops, 50–51
Expected value, 180
Experimental design, 273–282
mistakes in, 277, 278–279
terminology, 275–278
Experimental designs, types of, 279–282
fractional factorial, 18, 25, 281
full factorial, 18, 280, 343–367
half–replicate, 314
k factor, 381–389
one–factor, 327, 337–342
two–factor:
multiplicative models for, 353–359
with replications, 368–380
without replications, 343–367
2k factorial, 280, 283–292
2kr factorial, 293–313
22 factorial, 284–285
22r factorial, 293–308
2k–p fractional factorial, 314–326
Experimental errors, 222, 278
estimation of, 294–295, 346, 372
estimation of SSE, 329–330
standard deviation of, 228
test for independence, 234, 235, 237, 302, 324
Experimental unit, 277
Expert intuition, 421
Explicit instrumenting, 104
Exponential distribution, 488–489, 509, 510
truncated, 523–524, 531
use of, 488
variate generation, 475, 476, 489
Extended Fibonacci generators, 450
Extreme value distribution, 476
Factor analysis, 77
Factors, 17, 24, 77, 221, 276
principal, 77–81
selection of, 24
Fairness index, 36
FCFS (first come, first served), 509, 541, 551
F distribution, 489–490
approximation for, 638
percentiles of, 636
use of, 490
variate generation, 490
FEC (Flow–Equivalent Center), 611
Feedback shift register(s)
generalized, 449, 503
linear, 446
Fibonacci generators, 450
FIFO (first in, first out), 382, 383, 385
Figure of merit, 155, 157–158
File servers, 200, 568, 572–573
Filters, 104
Final conditions, 429
Finite buffer systems, 534–540
Firmware monitors, 100–101
First come, first served (FCFS), 509, 541, 551
Fixed–capacity service centers, 553, 570, 575, 576, 609
Flow balance, job, 551, 555
Flow control in computer networks, 592
Flow–Equivalent Center (FEC), 611
Flow law, forced, 557–560
FOM, see Figure of merit
Forced flow law, 557–560
Fork and join, 621
FORTRAN, as a simulation language, 402
Fpppp benchmark, 59
Fractile, 194
Fractional factorial designs, 25, 281, 314–326
Freedom, degrees of, 228, 332, 374, 489
F–test, 251, 267, 298, 333
Full factorial design, 280, 281
Function:
Ackermann’s, 50, 53
cumulative distribution, 180
probability density, 180
probability mass, 180
Games:
benchmarking, 130–131
decision makers’, 162
percentage, 169–170
pictorial, 146
ratio, 8, 165–174
winning, 170–173
Gamma distribution, 490–491
use of, 490
variate generation, 480, 490
Gantt charts, 149–152, 163
Garbage collection, 263, 311–313
GASP, 402, 403, 409, 502
Gateway response time, 527
Gateways, in networks, 524, 556, 568
Gaussian, see Normal distribution
GCC benchmark, 58
Generalized feedback shift register (GFSR), 449, 503
Generator polynomial, 320
Generators, load, 4
Geometric distribution(s), 200, 491
use of, 491
variate generation, 477, 491
Geometric mean, 187, 197, 271
GFSR (generalized feedback shift register), 449, 503
G/G/m queueing system, 544
Gibson instruction mix, 48, 49, 175
Global metrics, 34
G/M/1 queueing system, 543–544
Goals, 14–15
biased, 15
setting, 15
SMART, 42
Gordon and Newell’s method, 595
GPSS, 403, 409, 502
Grand mean, 328
Graphic charts, 140, 141–145, 176
checklist for, 143
guidelines for, 141
mistakes in, 163
Graphic displays, 418
Graphs:
double–whammy, 147
Kiviat, 153–159
Grid lines, 145
Half–replicate design, 314
Hanoi, towers of, benchmark, 167, 360, 361, 365
Hardware mapping, 98
Hardware monitors, 98
vs. software monitors, 98–100, 101
Harmonic mean, 188–189, 197
HB (higher is better) metric, 40, 153, 154, 168, 171, 172, 386
Heap, 410, 411
Hierarchical decomposition, 608, 613–620
Histogram, 144
broken scales in, 149
bucket (cell) sizes, 147
multiparameter, 76, 175
single–parameter, 75
Hit ratios, cache, 187
Homogeneity:
device, 552
routing, 552
Homoscedasticity, 237, 261
HP9000/840 processor, 354
Hybrid monitor, 101, 103, 175
Hyperexponential distribution, 509
Hypothesis testing, 213
iAPX 432, Intel, processor, 358, 359
IBM, 452
IBM PC, 69, 109
IBM PCjr, 69
Idle time, 39
IEEE, 8, 11
IFIP working group, 9
IID (independent and identically distributed), 241, 247, 461, 501, 507,
508, 516, 557
Implicit spying, 104
IMSL, 452
Independent and identically distributed (IID), see IID
Independent errors, test for, 234, 235, 237, 302
Independent events, 179
Independent job behavior, 552
Independent replications, 426–427, 431
Indexed linear list, 409
Indices of central tendencies, 183
Indices of dispersion, 193, 194, 195–196
Indirect binary n–cube networks, 433
Individual metrics, 34
Infinite servers (IS), 509, 551, 553
Informal methods, 386–389
Information systems, 242
INFORS, 10
Inherited time, 113
Initial conditions, 129, 395
Initial data deletion, 424–425
Initialization:
proper, 423–424
test for incorrect, 236
Input:
errors, 19
rate, 94
routine, 407
width, 94
Institute of Electrical and Electronic Engineers (IEEE), 8, 11
Instruction mix, 48–49
Gibson, 48, 49
Instrumentation mechanisms, 113
Instrumentation of programs, 175
Integration of systems, 132
Intel iAPX 432 processor, 358, 359
Intensity, traffic, 522
Interacting factors, 279
Interaction, 277, 279
Interactive response time law, 563
Interarrival times, 507
Interconnection networks:
comparison of, 289–290
Crossbar, 289–290
memory, 289–290
Omega, 289–290
simulation, 433
Intergroup variance, 87
Intermediate systems, 35
Internal driver, 132
International Federation for Information Processing (IFIP), 9
International Federation of Operational Research Societies (INFORS), 10
Interpretation:
of accounting logs, 117–119
of clustering results, 90
of data, 25, 102
of monitor results, 109
of normal quantile–quantile plots, 199, 200
Interquartile range, 194
Intragroup variance, 87
Intuition, expert, 421
Invalid models, 394–395
Inverse transformation, 474–476
I/O–arrow Kiviat graph, 155, 156
I/O–bound programs, 120
I/O optimization, 114
I/O–wedge Kiviat graph, 155
IS (infinite servers), 509, 551, 553
Jackson networks, 551
Job behavior, independent, 552
Job bounds, balanced, 585–591, 590
Job flow balance, 551, 555
Jobs, 510
classes of, 548, 551
in a system, distribution of, 593–595, 616, 619
Join and fork, 621
Journals, performance, 8, 9, 10
k–dimensional uniformity, 452, 467–470
k–distributivity, 452, 467–470
Keelboat Kiviat graph, CPU, 155
Kendall notation, 509
Kernels, 49–50, 175
2k experimental design, 280, 283–292
Keyboard emulation monitor, 176
k factor experimental designs, 381–389
Kilo Whetstone Instructions Per Second (KWIPS), 55
Kiviat graphs, 153–159, 176
for computer networks, 159
CPU–keelboat, 155
for databases, 159
I/O–arrow, 155, 156
I/O–wedge, 155
shapes of, 155–159
Knee, 565
capacity, 39, 125
Knowledge–based simulation, 503, 504
Kolmogorov–Smirnov test, see K–S test
2k–p experimental design, 314–326
sign table for, 316–317
2kr experimental design, 293–313
K–S (Kolmogorov–Smirnov) distribution, quantiles of, 637
K–S (Kolmogorov–Smirnov) test, 462–465
comparison with chi–square test, 465
KWIPS, 55
Lagrange multiplier technique, 241
Languages, for monitors, 97
LAN (local area network), 415, 621
Laplace distribution, variate generation, 480
Last Come, First Served (LCFS), with Preempt and Resume, 509, 551
LaTex, 322
Law(s):
Amdahl’s, 257, 258
forced flow, 557–560
Little’s, 512, 513–515, 560
operational, 555–568
definition, 556
response time:
general, 561–562
interactive, 563
of slow creation, 121
utilization, 556
Lawrence Livermore loops, 56
LB (lower is better) metric, 40, 153, 154, 168, 171, 172, 386
LCFS (Last Come, First Served), 509, 541
LCFS–PR (with Preempt and Resume), 509, 551
LCG, see Linear–congruential generators
Least–squares criterion, 223
Left–skewed pdf, see Skewed distribution
Legend box, 141
Length of simulation, 429–430
Level(s), 24, 255, 276
of accuracy, 32
of confidence, 204, 212–213
of detail, 18, 66–67, 394
of loading, 69
of significance, 204
LFSR (linear–feedback shift register), 446
Libraries, probe–point, 98
Limitations, of queueing theory, 620–622
Limit theorem, central, 182
Linear–congruential generators, 439–440
mixed, 440
multiplicative, 441–444
with m = 2k, 441
with m ≠ 2k, 441
Linear–feedback shift register, see LFSR
Linearizer algorithm, 624
Linear list, indexed, 409
Linear models, 400
Linear regression models, 221–243
multiple, 244, 245–250
simple, 221–243
Linear relationship, 266
test for, 235
Line chart, 140, 144, 145
misuse of, 145
Lines, grid, 145
Linked list:
doubly, 409
ordered, 409
UNPACK benchmark, 55
LISP interpreter benchmark, 58
Little’s law, 512, 513–515, 560
Live operators, 132
Livermore loops, Lawrence, 56
Load:
drivers, 132
generators, 4
Load–dependent:
arrivals, 621
service center, 554, 608–613
Loading level, 69, 77, 128
Local area network (LAN), 39, 415, 621
Local information, 552
Locality of reference, 120
Logic elements, 98
Logistic distribution, variate generation, 477
Lognormal distribution, 492
use of, 492
variate generation, 480, 492
Log transformation, 260–261
Long runs, 423
Loops, Lawrence Livermore, 56
Loss formula, Erlang’s, 536
Lottery, 212
Machine repairman model, 552
Macintosh, 69
Magnetic tape backup systems, 65–66
Majorizing function, 476
Management of capacity, 123
steps in, 124
Management of systems, 103
Manager, system, 103
Many–to–many mapping, 103
Mapping hardware, 98
Markoff, see Markov entries
Markov chain, 516
Markov models, 81–82
Markov processes, 516
Mass function, probability, 180
Matrix:
benchmark, 58
transition, 81–82
M/D/1 queue, 542–543
Mean, 180, 183
arithmetic, 73, 188, 197
confidence interval for, 204–207
geometric, 187, 191, 197
harmonic, 188–189, 197
misuses of, 186
of ratio, 189–192
relationship to median, mode, 183–186
sample size for determining, 216–217
selecting median, mode, and, 183–186
standard error of, 205, 219
of sums, 181
testing for zero, 207–208, 216
Mean absolute deviation, 193, 194, 197
Mean arithmetic, 305
Mean queue length, 429–430, 602
Mean square of errors, 251, 299. See also MSE (mean squared error)
Mean time to failure (MTTF), 40
Mean–value analysis, 570, 575–579
approximate, 579–585
with load–dependent centers, 608–613
Schweitzer’s approximation, 580, 581
Mean waiting time, 429
Measure, when to, 30
Measured data:
for simulation validation, 422
summarizing, 179, 182–183
validation of, 32, 129
Measurement techniques, program, 113
Median, 181, 197
relationship to mean, mode, 183–186
selecting mean, mode, and, 183–186
Memoryless distribution, 488–489, 509
Memory management, dynamic, 408
Memory management systems, 311, 622
Method of subsamples, see Batch means
Metrics, 4, 23
commonly used, 36
for computer networks, 159
distance, 86
global, 34
HB (higher is better), 40, 153, 154, 168, 386
individual, 34
LB (lower is better), 40, 153, 154, 168, 386
mistake in selecting, 16
NB (nominal is best), 40
selection, 33–37, 61
utility classification of, 40
MFLOPS, 38, 49, 56
M/G/1 queue, 540–541
M/G/∞ queue, 543
Millions of Floating–point Operations Per Second, see MFLOPS
Millions of Instructions per Second, see MIPS
Minimal standard random number generator, 452
Minimum spanning tree method, 87
MIPS, 16, 38, 49
averaging, 188
Missing observations, 360–366
Miss ratios, cache, 187
Mistakes:
in benchmarking, 127–130
in charts, 144–145
in evaluation, 14–22
how to avoid, 14–29
in experimental design, 277, 278–279
in model validation, 583
in queueing approximations, 583
in regression, 266–269
in selecting metrics, 16
in simulation, 393–397
Misuses:
of line charts, 145
of means, 186
of percentages, 170
of statistics, 271
Mixed linear–congruential generators, 440
Mixed queueing networks, 548
M/M/m/B queue, 534–540
M/M/m queue, 527–534
M/M/∞ queue, 540
M/M/1 queue, 522–526
Mode, 181, 197
relationship to mean, median, 183–186
selecting mean, median, and, 183–186
Modeling error, see Error
Model parameter estimation, 223–226, 284–285, 294, 328, 344–346, 370
Model(s):
additive, 303–306
central server, 552
definition of a good, 222–223
deterministic, 400, 417
for experimental designs, 327, 344–346, 368–369, 381–382
linear and nonlinear, 400
linear regression, 221–243
multiple, 244, 245–250
simple, 221–243
machine–repairman, 552
Markov, 81–82
multiplicative, 303–308, 353–359
open and closed, 400–401
probabilistic, 400
queueing network, 553
stable and unstable, 401
static and dynamic, 400
stochastic, 10
Model validation, 413–414
mistakes in, 583
techniques, 420
Model verification, 413–414
Modular design, 414–416
Modularity, 414
Modulus, 439
Monitor(s), 4, 93–110, 175
accounting logs as, 114, 116–121
activation mechanisms for, 95–96
advertising by, 105
buffer size in, 96
data collection in, 105
priority, 97
soliciting by, 105
terminology, 94
types of, 94–95
commercially available, 175
distributed system, 101–109
hardware, 98
keyboard emulation, 176
program execution, 111–114
software, 95, 175
Monitoring:
abnormal events, 97–98
on–off switch for, 97
overhead, 129
reasons for, 93
Monitor languages, 97
Monte Carlo simulation, 403
Moving average method, 427
MSA (Mean Square of A), 333
MS–DOS, 70, 282
MSE (mean squared error), 229, 246, 251, 299, 333, 348
MSR (mean squared residual), 246, 251
MTTF, see Mean time to failure
Multicollinearity, 253–254
Multiparameter histograms, 76, 175
Multiple correlation coefficient, 246, 249
Multiple linear regression, 244, 245–250, 266. See also Linear regression models
Multiplicative linear–congruential generator, 440–444
with m ≠ 2k, 441
Multiplicative models
for 22r experiments, 303–308
for two–factor experiments, 353–359
Multiplier, 439
Multiprocessor systems, 527
Multistream simulations, 454
Myths, random number generation, 455–458
Name server, 518
NASA7 benchmark, 58
NB (nominal is best) metric, 40
Negative binomial distribution, 492–493
use of, 486, 492–493
Nominal capacity, 38, 125
Nondistributed systems, 101
Nonexecutable workload, 67
Nonexponential service times, 620–621
Noninteracting factors, 279
Nonlinear models, 400
Nonproduct form networks, 620
Nonredundancy of metrics, 35
Nonregenerative simulation, 434
Nonzero origins in charts, 146
Normal distribution, 181–182, 493–494, 499, 500
approximation for, 198–199, 638
area under, 628, 638
quantiles of, 629, 630
standard, 181–182, 493
test for, 237, 302
unit, 182, 493
use of, 494
variate generation, 480, 494
Normalization (for clustering):
percentile, 86
range, 85–86
variance, 85
weight, 85
Normalizing constant, 603
Normal quantile–quantile plot, 198, 237, 303, 305, 307, 335, 350, 356, 363, 364, 366, 378, 385
interpretation of, 199, 200
Norton’s theorem, 614
Notices, event, 408
Number of jobs in a queue, 511
vs. in system, 511
vs. system, distribution of, 619
vs. time in system, 512
Observation method, 386–387
Observations:
correlated, 430–431
missing, 360–366
in monitors, 104–105
paired, 209
system, 102
unpaired, 209–210
Observer, 102
Omega interconnection networks, 289–290
Omega transformation, 261
One–factor–at–a–time designs, 279
One–factor experimental designs
with equal sample sizes, 327–341
ANOVA table for, 333
with unequal sample sizes, 337–342
ANOVA table for, 338, 340
One–sided confidence intervals, 213–215
for means, 219
for proportions, 219
One–step behavior, 551
On–line graphic displays, 418
On–line monitors, 95
Open models, 400–401
Open queueing networks, 547–548, 551
analysis of, 570, 572, 575–579
Operating systems:
ARGUS, 230–232, 239, 240, 256
CPM, 282
MS–DOS, 70, 282
PRIMOS, 452
UNIX, xii, 230, 231, 256, 282, 359
VMS, 359
WANG VS, 324
Operational laws, 555–568
definition, 556
Operationally testable assumptions, 555
Operational quantities, 555
Operations Research Society of America (ORSA), 10
Operators, live, 132
Optimization:
of code, 114, 120
of I/O, 114
of paging, 114
Optimizer, program, 192
Ordered linked list, 409
Ordered qualitative variable, 140
Organizations, professional, 8
ORSA, 10
Orthogonal vectors, 77, 288
Outliers, 19, 83, 85, 183, 193, 194, 265
Output analysis techniques for simulations, 502
Overflow counter, 97
Overflow of buffers in monitors, 97
Overhead, monitor, 94
Overlapping pain, 468–472
Page faults, 120
Paging
algorithms, 382, 404
optimization, 114
performance, 383–385
study, 382–385, 389
Paired observations, 209
P2 algorithms for percentiles, 196
Parallelism
in space, 354
in time, 354
Parallel simulation, 504
Parameters, 17, 23, 204–207
confidence intervals for, 229–232, 240, 241, 246, 249, 298–299, 335–337, 351–353, 374–379
estimation, 223–226, 284–285, 294, 328, 346, 370
regression, 223
validation, 421
for workloads, 83
Pareto distribution, 201, 495
use of, 495
variate generation, 477, 495
Partial factorial designs, 18
Pascal, 359
program for random number generation, 443
program for sieve benchmark, 53
routines for simulation, 503
as a simulation language, 402
Pascal distribution, 495–496
use of, 495
variate generation, 480, 496
pdf, see Probability density function (pdf), skewed
PDF, see Probability distribution function
PDP–11/70 processor, xii, 167, 361, 365
Percentages, ratio games with, 169–170
Percentile normalization, 86
Percentiles, 194, 195, 197
Performance:
conferences, 8
evaluation, see Evaluation
journals, 8–9, 10
projects, 11
requirements, setting, 40–43
tuning, 123
Performance evaluation
art of, 7
mistakes in, 14–29
Performance Evaluation Review, 8
Performance monitoring, 108
Period of generators, 438
Pictograms, 148
Pictorial games, 146–149
Pie chart, 144, 146
Pipes, remote, 25–27
Plotting random quantities, 147–148
pmf, see Probability mass function
Poisson arrivals, 508
bulk, 509
Poisson distributed variable, 261
Poisson distribution, 201, 496
use of, 486, 493, 496
variate generation, 480, 496
Poisson processes/streams, 201, 496, 516–517
properties of, 517
Polar method, 494
Pollaczek–Khinchin (P–K) mean value formula, 541
Polynomial:
characteristic, 445
generator, 320
primitive, 445, 468
Population size, 508
Possession, single resource, 552
Power, metric, 36
Power transformation, 262
PPS (packets per second), 38, 524
Predictability of random numbers, 456
Predicted responses, confidence intervals for, 232–234, 240, 241, 246, 250, 299–301
Prediction beyond measured range, 268
Predictors, 221, 276
categorical, 254–257
correlated, 268
too many, 269–269
Preempt and Resume, Last Come First Served, 509, 551
Presentation of data, 20, 25, 102, 106–108, 139–163
Price–performance ratio, 58
Primary factors, 276
Primitive polynomials, 445, 449, 459, 468
Primitive trinomials, 449
PRIMOS operating system, 452
Principal–component analysis, 76–80, 175
Principal factor, 77
Print server, 568
Probabilistic models, 400
Probability density function (pdf), 180, 181. See also Distributions; particular distributions
skewed, 186
Probability distribution function (PDF), 180, 181. See also Distributions; particular distributions
Probability mass function (pmf), 180. See also Distributions; particular distributions
Probability theory, 179
Probe points, 98, 104
Probing, 104
Problem(s):
in capacity planning, 125
with clustering, 90–91, 175
of multicollinearity, 253–254
Procedure call, remote, 26, 230–232
Processes:
arrival, 507–508, 551
birth–death, 516, 519–522
continuous–state, 515–516
discrete state, 515–516
Markov, 516
Poisson, 201, 496, 516–517
relationship among, 517
stochastic, 515–517
types of, 515–517
Processing kernel, 50
Z80 Processor, xii, 275, 276, 279, 281
Z8002 Processor, 167, 360, 361, 365
68000 Processor, xii, 275, 276, 281, 282, 359, 360, 361, 365, 367
8080 Processor, 166, 276
8086 Processor, xii, 275, 281, 282, 359
80286 Processor, 282
Processor comparison study, 357
Processors:
68000, xii, 275, 276, 281, 282, 359, 360, 361, 365, 367
8080, 166, 276
8086, xii, 275, 281, 282, 359
80286, 282
C/70, 167, 360, 361, 365
HP9000/840, 354
Intel iAPX 432, 358, 359
PDP–11/70, xii, 167, 360, 361, 365
RISC–I, 167, 360, 361–365
VAX–11/780, xii, 167, 359, 360, 361, 365
Z80, xii, 275, 276, 279, 281
6502, xxvii, 166
Z8002, 167, 360, 361, 365
Processor sharing, 509, 551
Product–form networks, 548–552
Productivity metrics, 34
Professional organizations, see Organizations, professional
Profile:
execution, 113
resource usage, 68
Profile report, 113
Program(s):
efficiency of, 115, 120
execution monitors, 111–114 design, 112
instrumentation, 113, 175
measurement techniques, 113
optimizer, 191, 192
performance improvement, 114, 115–116
synthetic, 50–51
text formatting, 322
usage in educational environments, 74
Program execution, monitor design, 112
Projection validation, 126
Projects, performance, 11
Proportions, 215
confidence intervals for, 215–216
sample size for determining, 217–218
Pseudo–random numbers, 438
PS (processor sharing), 551
Puzzle benchmark, 50, 167, 359, 360, 361, 365
QNET4, 402
Qualitative variables, 140
ordered, 140
unordered, 140
Quantile(s), 181. See also Percentiles
of chi–square distribution, 632–633, 638
of F distribution, 634, 638
of K–S (Kolmogorov–Smirnov) distribution, 637
of normal distribution, 198, 629, 630, 638
of t distribution, 631
Quantile–quantile plots, 198, 202, 271
Quantitative variables, 140
continuous, 140
discrete, 140
Quartiles, 194
Queue(s), 509
analysis of, 519–546
calendar, 410
components of, 508
G/G/m, 544
G/M/1, 543–544
M/D/1, 542–543
M/G/∞, 543
M/G/1, 540–541
M/M/∞, 540
M/M/1, 522–526
M/M/1/B, 533, 539
M/M/m, 527–534
M/M/m/B, 534–540
M/M/m/B/K, 546
M/M/m/K, 546
notation, 507–510
rules for all, 509
Queueing analysis, symbols, 624–626
Queueing approximations, mistakes in, 583
Queueing network(s), 547–554
BCMP, 551
closed, 547–548
large, 608–620
mixed, 548
models of computer systems, 553
nonproduct form, 620
open, 547–548, 551
analysis of, 570, 572, 575–579
product form, 548–552
Queueing notation, 507–510
Queueing theory:
introduction to, 507–518
limitations of, 620–622
Queue length, 511
distribution of, 599–600, 601, 609, 619
mean computation, 430
Random number, 82
Random–number generation, 437–458, 502, 503
nonoverlapping streams, myths, 454
Random–number generators
combined, 451–453
desired properties of, 437–439
Fibonacci, 450
implementing, 456
linear–congruential, 439–440
Pascal code for, 443
myths about, 455–458
poor, 395
RANDU, 452
survey of, 452–453
Tausworthe, 444–450
testing, 456, 460–472, 502
in UNIX, 453
Random numbers, 395
bits of, 457
predictability of, 456
nonoverlapping pairs, 468, 469
nonoverlapping streams, 454
Random variable, 179
Random–variate generation, 474–482. See also Distributions; particular types of distributions
technique, finding, 481
RANDU, 452
Range, 193, 197, 389
Range method, 388–389
Range normalization, 85–86
Ranking method, 387
Rat holes, 162
Ratio(s):
cost/performance, 40
hit, 187
mean of, 189–192
miss, 187
reads to writes. 121
visit, 558, 560
Ratio games, 8, 165–174
with percentages, 169–170
relative performance enhancement, 168
strategies for winning, 170–173
Ratio metrics, 35, 168
Reaction time, 38
Reads to writes, ratio of, 121
Real measurements, 422
Real variables, 140
Real workload, 47, 128
Reduced Instruction Set Computer, see RISC
Reduction techniques, variance, 436, 502
Reference, locality of, 120
Regeneration, 433–435
cycle, 434
method of, 434
point, 434
Regenerative simulations, 502
Regenerative system, 434
Regression analysis, 221, 244–271
assumptions, 234
visual tests for, 234–241, 266
checklist for, 269–270
mistakes in, 266–269
Regression models, 221–243, 244–271
with categorical predictors, 254–257
curvilinear, 257
good and bad, 222
linear, 221, 241
multiple, 244, 245–250
simple, 221
other, 244–271
over partial range, 268–269
predictions, confidence intervals for, 232–234
transformations for, 259
Regression parameters, 223
confidence intervals for, 229–232, 267
estimation of, 223–226, 241, 246
numerical values of, 266
Regression testing, 132
Rejection method, 476, 494
Reliability, 34, 40, 42
Remote pipes, 25–27
Remote procedure call, 26, 230–232, 239, 256
Remote–terminal emulation, 132–133
Remote–terminal emulator (RTE), 4, 132–138, 176
components emulated by, 133
components of, 134
output, 135
Repairman model, machine, 552
Repeatability of workloads, 69
Repetitions, 407
Replacement algorithm, 382
Replication(s), 276
2kr factorial designs with, 294
moving average of independent, 426–427
two–factor designs with, 368–380
Report, profile, 113
Report generator, 408
Representativeness, 67
Requirements, setting, 40–43
Reservation systems, airline, 570
Residual, see Error
Resolution:
design, 321–326
monitor, 94
Resource possession:
simultaneous, 622
single, 552
Resource usage profile, 68
Response(s), 276
Response–dependent arrivals, 621
Response time, 4, 30, 37, 601, 616
variability of, 36
Response time law:
general, 561–562
interactive, 563
Response variable, 221, 276
confidence intervals for, 232–234, 250, 299–301
Responsiveness, 34
resq, 402
Results:
presentation of, 20
reasons for not accepting, 161–162
saleability of, 32
22r experimental designs, 293–308
multiplicative model for, 303–308
RISC, 12, 16, 354–356
RISC architecture (Scheme and Spectrum), 354–356
RISC–I code size study, 167, 329, 365, 390
RISC–I processor, 166, 167, 360, 361, 365
Round–robin (RR), 509
Routines
event, 407
input, 407
trace, 408
Routing homogeneity, 552
RPC (remote procedure call), 26, 230–232, 239, 256
RTE, see Remote–terminal emulator
Run length, 432
Runs, long, 423
Saleability of results, 32
Sample, 203–207
mean, 183
median, 183
mode, 183
standard deviation, 193, 197
vs. population, 203–204
Sample data:
comparing systems using, 203
summarizing, 179
Sample size:
for comparing two alternatives, 18
for determining mean, 216–217
for determining proportions, 217–218
Sample variance, 193, 197
Sampling, 84, 113, 128
Sampling monitors, 95
SAS statistical package, 453
Scale:
in histograms, broken, 149
ranges, selection of, 145
Scaling, data, 85
Scaling factor, 594, 595
Scatter diagram(s), 76, 77, 224, 230, 231, 235, 236, 237, 238, 257, 260, 267, 303, 335, 350, 469
Scatter plots, see Scatter diagram(s)
Scenario, 133
Scheduler design study, 325, 387, 390
Scheduling:
CPU, 404
simulation of, 407, 423–424, 428–430, 433
events, 503
Scheme RISC architecture, 354–356
Schrage’s method, 442, 444
Schumacher charts, 160, 176
Schweitzer’s approximation, 580, 581, 583
Script, 133
SCS, see Society for Computer Simulations
Secondary factors, 276
Seed(s), 395, 438
independence, 420
selection of, 395, 453–455, 456
table of, 455
Self–time, 113
Semi–Interquartile Range (SIQR), 194–195, 197
Sequence:
Fibonacci, 450
Tausworthe, 444, 446
Serial–correlation test, 465
Serial test, 468–470
Server(s):
central, 552
file, 200, 568, 572–573
infinite, 551
print, 568
Service center(s), 509
fixed–capacity, 553, 570, 575, 576, 577, 609
flow equivalent, 611
load–dependent, 554, 608, 609, 610
Service discipline(s), 509, 551
fair, 552
first come, first served (FCFS), 509, 541, 551
last come, first served (LCFS), 509
with preempt and resume (LCFS–PR), 509, 551
processor sharing (PS), 551
in random order (SIRO), 541
Services, 23, 26, 62
Service times:
distribution, 509, 551
nonexponential, 620–621
state–dependent, 551
Shift register(s):
generalized feedback, 448–449, 503
linear feedback, 446
Shorted model, 614, 615, 618
SIAM, 9, 10
Sieve benchmark, 50, 52, 175, 345, 346, 357, 359
Pascal listing for, 53
SIGMETRICS, 8
Signaling theory, 262
Significance level, 204–207
Sign table, 286, 292
for 2k–p experimental design, 316–317
SIGSIM, 9
Simple designs, 18, 279–280
Simple linear regression, 221
SIMPL/I system, 452
SIMSCRIPT, 402, 403, 452, 502
SIMULA, 402, 403, 409, 416, 453, 502
listing of Ackermann’s function, 53
Simulation:
of computer networks, 414–415
of CPU scheduling, 433
types of, 403–408
concurrent, 503
continuous, 403
discrete event, 403, 406–408
emulation, 403
event driven, 403, 407
knowledge–based, 503, 504
Monte Carlo, 403
multistream, 454
nonregenerative, 434
parallel, 503
regenerative, 434, 502
static, 403–404
terminating, 428–430
trace driven, 404–406, 502
Simulation analysis, 393–412, 413–436
failure, causes of, 395–397
mistakes in, 393–395
when to use, 31
Simulation books, 502–503
Simulation clock, 407
Simulation language(s), 401–403, 416, 502
CSMP, 403
DYNAMO, 403
FORTRAN as a, 402
GASP, 402, 403, 409, 502
GPSS, 402, 409, 502
Pascal as a, 402
selection of, 401–403
SIMSCRIPT, 402, 403, 452, 502
SIMULA, 402, 403, 409, 453, 502
Simulation length, 395, 430
Simulation output analysis, 413, 502
Simulation packages, 402
QNET4, 402
RESQ, 402
Simulation team, 397
Simultaneous resource possession, 622
Sine transformation, arc, 261
Single–parameter histogram, 75
Single–resource possession, 552
SIQR (Semi–Interquartile Range), 194–195, 197
SIRO (service in, random order), 541
6502 processor, xxvii, 166
68000 processor, xii, 275, 276, 280, 359, 360, 361, 366, 367
Sizing, system, 124–125
Skewed distribution, 183, 185, 186
left–, 183
right–, 183
Skewness, 185
coefficient of, 197
of device demands, 128
Skills:
for simulation, 397
social, 20–21
substantive, 20–21
Slow creation, law of, 121
SMART goals, 42, 396
Social skills, 20–21
Society for Computer Simulations (SCS), 9, 503
Society for Industrial and Applied Mathematics, see SIAM
Software monitors, 95, 175
vs. hardware monitors, 98–100, 101
Soliciting, by monitors, 105
Spanning tree method, minimum, 87
SPEC benchmark suite, 58
SPECmark, 59
SPECthruput, 59
Spectral test, 470–472
Spectrum RISC architecture, 354–356
Speed, 34, 42
Spice benchmark, 58
Square criterion, least, 223
Square–root transformation, 259
Squares, sum of, 226, 287
SSE (sum of squared errors), 223, 225, 228, 295, 296, 297, 299, 331–332, 348, 373
SS0, 227, 297, 331–332, 348, 373
SSR, 227
SST (sum of squares total), 226–227, 287, 296, 297, 331–332, 348, 373
SSY, 227, 297, 331–332, 348, 373
Stability condition, 511
Standard deviation, 180, 193, 197
vs. hardware monitors
errors, 228
test for constant, 238
of parameters, 246
Standard error, 205
of sample mean, 219
Standard normal distribution, 181
State:
continuous, 399
discrete, 399
steady, 129–130, 395, 401, 413, 423, 428, 520, 621
transient, 129–130, 423
State–dependent service, 551
State probabilities, 595, 604, 607
State transition diagram, 81
for birth–death processes, 520
for M/M/m/B queues, 535
for M/M/m queues, 527
for M/M/1 queues, 522
Static models, 400
Static simulation, 403–404
Statistical package, SAS, 453
Statistical tables, 627–638
approximation formulas for, 638
Statistics, 204–207
basic concepts, 179
misuse of, 271
Steady state, 395, 520
Steady–state performance, 423
Steady–state probability, of birth–death
process, 520
Stochastic model, 10
Stochastic processes, 515–517
relationship among, 518
Stopping criteria, 430–431
Storage, dynamic allocation of, 404
Storage systems, benchmarking, 176
Strategies, ratio game, 170–173
Streams:
nonoverlapping, 454
Poisson, 201
Stress–load analysis, 132
Stretch factor, 38
Structured walk–through, 416–417
Student’s t distribution, see t distribution
Subsample, 427
Substantive skills, 20–21
data, 179, 182, 197
variability, 192
Sum of squared errors, see SSE
Sum of squares total, see SST
SUT (system under test), 61, 62, 133
Symbols, queueing analysis, 624–626
Symbols in charts, 144
Synthetic programs, 50–51
Synthetic workload, 47
System, 4
balanced, 585
end, 35
finite buffer, 534–540
finite population, 546
intermediate, 35
vs. queue, 511, 512
regenerative, 434
System capacity, 509
System console, 102–103
System crashes, time between, 214
System integration, 132
System management, 103
System observation, 102
System parameters, 17, 23
System performance evaluation cooperative (SPEC), 58
Systems:
airline reservation, 570
banking, 56–57, 570
batch, 594
comparison of, 203–220
several, 327
two, 18, 208–209
database, 6, 12, 201, 242
distributed, 101–109, 200, 568
information, 242
magnetic tape, 65–66
memory management, 311, 622
multiprocessor, 40, 527
storage, 176
text formatting, 322
timesharing, 62, 568, 602–607
transaction processing, 569, 570, 591
word processing, 388
System sizing, 124–125
System under test (SUT), 61, 62, 133
Table(s):
seeds, 455
sign–, 286, 292, 316–317
statistical, 627–638
approximation formulas for, 638
Taguchi method, 390
Tail, 438
Tape backup system, magnetic, 65–66
Tausworthe generators, 444–450
disadvantage of, 449
Tausworthe sequence, 468
t distribution, 497
quantiles of, 631
use of, 497
variate generation, 497
Terminating simulations, 428–430
Terminology:
experimental design, 275–278
workload characterization, 71–90
Test(s):
chi–square, 461–462
consistency, 420
for constant variance, 237, 302
continuity, 418–419
for correct initialization, 236
degeneracy, 419
F, 251, 298, 333
hypothesis, 213
for independent errors, 235, 302, 324
K–S (Kolmogorov–Smirnov), 462–465
for linear relationship, 235
for normality, 237, 302
for random–number generators, 456, 460
serial, 468–470
serial–correlation, 465
spectral, 470–472
statistical, 271
t–test, 209–211
two–level, 466–467
for zero mean, 207–208, 216
Testing:
hypothesis, 213
regression, 132
Test workload, 47, 128, 175
Text formatting programs, 322
Theorems:
central limit, 182
Chandy–Herzog–Woo, 615
Norton’s, 614
Think time, 134, 553, 622
Three–quarter high rule, 146
Throughput, 38, 616
Time:
inherited, 113
vs. number in queue, 512–513
parallelism in, 354
between system crashes, 214
Tune–advancing mechanisms, 407, 503
Time in queue, vs. time in system, 512–513
Timeliness, 68
Time–rate–resource metrics, 34
Timer interrupt, 96
Timesharing systems, 62, 568, 602–607
Time to failure, 40
Tomcatv benchmark, 59
Top–down design, 414–416
Towers of Hanoi benchmark, 167, 360, 361, 367
TPC benchmark, xii, 58
TPC (Transaction Processing Performance Council), xii, 58
TPS (transactions per second), 4, 38, 57
Trace, 67, 94, 404, 417
event, 418
mode, 96
routines, 408
Trace–driven simulations, 404–406, 502
advantages of, 405
disadvantages of, 406
Tracing, 111, 113
Trade–off evaluation, 31
Traffic intensity, 522
Train arrivals, 621
Transaction Processing Performance Council
(TPC), xii, 58
Transaction processing systems, 569, 570, 591
Transactions per second, see TPS
Transformation(s), 83, 84, 259–265, 271
determining, 260
inverse, 474–476
log, 260–261
omega, 261
power, 262
square–root, 259, 261
to stabilize the variance, 262
Transient:
performance, 129–130
removal, 423
Transient analysis, 621
Transient state, 130, 423
Transition probability matrix, 81, 82, 559
Trap instruction, 96
Treatment, 276
Tree, minimum spanning, 87
Tree structures, 410, 411
Triangular distribution, 480
troff, 322
TRS–80, 109
Truncated exponential distribution, 523–524, 531
Truncation method, 424
t–test:
paired, 209
unpaired, 210
Tuning, performance, 111, 123
Turnaround time, 38
Two–factor experimental designs
multiplicative models for, 353–359
with replications, 368–380
ANOVA table, 374
without replications, 343–367
ANOVA table, 349
2k experimental designs, 280, 283–292
2k–p experimental designs, 314–326
2kr experimental designs, 293–313
22r experimental designs, 293, 309
Two–level tests, 466–467
Uniform distribution (discrete)
use of, 498
variate generation, 498
Uniformity, k–dimensional, 467–470
Unimodal distributions, 183
Unit–time approach, 407
UNIX operating system, xii, 230–232, 256, 282, 359
random number generator, 453
Unordered qualitative variables. 140
Unpaired observations, 209–210
Unrepresentative workload, 16
Unstable model, 401
Unverified models, 394
Uptime, 40
Usable capacity, 39, 125
User participation in analysis, 397
Users, 125
Utility classification of metrics, 40
Utilization, 34, 39, 601, 602, 603, 606, 616
law, 556
use for validation, 130
Validation
of Amdahl’s law, 257, 258
of assumptions, 421
baseline, 126
of conclusions, 421
of measured data, 32, 130
of parameters, 421
projection, 126
rules for, 32
of simulation models, 420–421
using utilization, 130
Variability, 19
of metrics, 35
of response time, 36
summarizing, 192
Variables, types of, 140
continuous, 140
discrete, 140
predictor, 221, 269
qualitative, 140
quantitative, 140
random, 179
response, 221
Variance, 180, 193, 197, 287
intergroup, 87
intragroup, 87
of queue length, 593
of sample mean, 181, 430–431
of sums, 181
test for constant, 302
Variance estimation, 430–431
Variance normalization, 85
Variance reduction, 436, 502, 503
Variate, 437
Variate generation, 474–482
Variation, 226, 286–291
allocation of, 226–228, 295–298, 331–332, 347–348, 372–374
coefficient of, 181, 197
difference from variance, 287
VAX–11/780 processor, xii, 167, 359, 365
Vectors, orthogonal, 77
Verification, of models, 413–414
Visit ratio, 558, 560
Visual tests:
for confidence intervals, 211–212
for equality, 211–212
for experimental designs, 302–303, 309, 334, 366, 379
for regression, 234–241, 266
VMS operating system, 359
Waiting time, mean, 429
Walk–through, structured, 416–417
WANG VS operating system, 324
Weibull distribution, 498
use of, 498
variate generation, 477, 499
Weighted Euclidean distance, see Euclidean distance
Weight normalization, 85
Whetstone benchmark, 54, 55
Word processing systems, 388
Workload, 4, 24
characterization, 71–90, 175
using accounting logs, 121
using average values, 74
component, 72
for computer networks, 63–65
features, 72
loading level of, 69
parameters, 17, 23, 72
real, 128
selection, 60–69
test, 128
types of, 47, 48–52
Workload unit, 72, 125
Workstation
design, 275–276
emulation, 138, 176
Writes, ratio of reads to, 121
Zero mean, testing for, 207–208
Z80 processor, xii, 275, 276, 279, 281
6502 processors, 166
Z8002 processor, 167, 360, 361, 365


Previous Table of Contents Next

Copyright © John Wiley & Sons, Inc.