Software Performance and Scalability: A Quantitative Approach

Software Performance and Scalability......Page 5
Contents......Page 9
PREFACE......Page 17
ACKNOWLEDGMENTS......Page 23
Performance versus Scalability......Page 25
PART 1 THE BASICS......Page 27
1. Hardware Platform......Page 29
1.1 Turing Machine......Page 30
1.2 von Neumann Machine......Page 31
1.3 Zuse Machine......Page 32
1.4.2 Hyperthreading......Page 33
1.4.3 Intel’s Multicore Microarchitecture......Page 37
1.5 Sun Machine......Page 41
1.6.1 Processors......Page 42
1.6.2 Motherboard......Page 43
1.6.3 Chipset......Page 44
1.6.4 Storage......Page 46
1.6.5 RAID......Page 48
1.6.6 Networking......Page 51
1.6.7 Operating System......Page 53
1.7.1 Memory Leaks......Page 54
1.8 Sizing Hardware......Page 59
Recommended Reading......Page 61
Exercises......Page 62
2. Software Platform......Page 65
2.1 Software Stack......Page 66
2.2 APIs......Page 68
2.2.2 Java APIs......Page 69
2.2.3 Google APIs......Page 70
2.3 Multithreading......Page 71
2.4.1 Systems Software......Page 77
2.4.2 Application Software......Page 78
2.5.1 What Is Enterprise Software?......Page 79
2.5.3 Monolithic Architecture......Page 81
2.5.4 Client/Server Architecture......Page 82
2.5.5 Three-Tier Architecture......Page 83
2.5.6 N-Tier Architecture......Page 84
2.5.8 Service-Oriented Architecture......Page 85
2.6 Summary......Page 87
Exercises......Page 88
3. Testing Software Performance and Scalability......Page 89
3.1 Scope of Software Performance and Scalability Testing......Page 91
3.1.1 Performance Regression Testing......Page 92
3.1.2 Performance Optimization and Tuning Testing......Page 94
3.1.4 Scalability Testing......Page 99
3.1.6 Additional Merits of Performance Testing......Page 106
3.2.1 Agile Software Development......Page 107
3.2.2 Extreme Programming......Page 108
3.3 Defining Software Performance......Page 110
3.3.1 Performance Metrics for OLTP Workloads......Page 111
3.3.2 Performance Metrics for Batch Jobs......Page 116
3.4 Stochastic Nature of Software Performance Measurements......Page 119
3.5 Amdahl’s Law......Page 121
3.6 Software Performance and Scalability Factors......Page 123
3.6.1 Hardware......Page 124
3.6.2 Operating System......Page 127
3.6.3 Database Statistics......Page 131
3.6.4 SQL Server Parameterization......Page 132
3.6.6 Licensing......Page 134
3.7 System Performance Counters......Page 135
3.7.1 Windows Performance Console......Page 136
3.7.2 Using perfmon to Diagnose Memory Leaks......Page 142
3.7.3 Using perfmon to Diagnose CPU Bottlenecks......Page 143
3.7.4 Using perfmon to Diagnose Disk I/O Bottlenecks......Page 145
3.7.5 Using Task Manager to Diagnose System Bottlenecks......Page 149
3.7.6 UNIX Platforms......Page 152
3.8 Software Performance Data Principles......Page 153
3.9 Summary......Page 155
Recommended Reading......Page 156
Exercises......Page 157
PART 2 APPLYING QUEUING THEORY......Page 159
4. Introduction to Queuing Theory......Page 161
4.1 Queuing Concepts and Metrics......Page 163
4.1.1 Basic Concepts of Queuing Theory......Page 164
4.1.2 Queuing Theory: From Textual Description to Mathematical Symbols......Page 165
4.2.1 Random Variables and Distribution Functions......Page 167
4.2.2 Discrete Distribution and Probability Distribution Series......Page 168
4.3 Applying Probability Theory to Queuing Systems......Page 169
4.3.1 Markov Process......Page 170
4.3.2 Poisson Distribution......Page 172
4.3.3 Exponential Distribution Function......Page 174
4.3.5 Queuing Node versus Queuing System......Page 176
4.4 Queuing Models for Networked Queuing Systems......Page 177
4.4.1 Queuing Theory Triad I: Response Time, Throughput, and Queue Length (Little’s Law)......Page 178
4.4.2 M/M/1 Model (Open)......Page 179
4.4.4 Queuing Theory Triad II: Utilization, Service Time, and Response Time......Page 183
4.4.5 Multiple Parallel Queues versus Single-Queue Multiple Servers......Page 184
4.4.6 M/M/m/N/N Model (Closed)......Page 186
4.4.7 Finite Response Time in Reality......Page 190
4.4.8 Validity of Open Models......Page 193
4.4.9 Performance and Scalability Bottlenecks in a Software System......Page 194
4.4.10 Genealogy of Queuing Models......Page 195
4.5 Summary......Page 196
Recommended Reading......Page 198
Exercises......Page 199
5. Case Study I: Queuing Theory Applied to SOA......Page 201
5.1 Introduction to SOA......Page 202
5.2 XML Web Services......Page 203
5.3 The Analytical Model......Page 205
5.4 Service Demand......Page 207
5.4.2 XML SOAP Serialization/Deserialization......Page 208
5.4.3 Network Latency......Page 209
5.4.5 Database Server......Page 210
5.4.6 Data Storage......Page 211
5.5.1 Exposing a Stateless Session EJB as an XML Web Service......Page 212
5.6 MedRec Deployment and Test Scenario......Page 213
5.7 Test Results......Page 215
5.7.1 Overhead of the XML Web Services Handle......Page 216
5.7.2 Effects of Caching Web Services Handle......Page 217
5.7.3 Throughput Dynamics......Page 218
5.7.4 Bottleneck Analysis......Page 219
5.8 Comparing the Model with the Measurements......Page 222
5.10 Summary......Page 224
Recommended Reading......Page 225
Exercises......Page 226
6. Case Study II: Queuing Theory Applied to Optimizing and Tuning Software Performance and Scalability......Page 229
6.1.1 Characterizing Performance and Scalability Problems......Page 231
6.1.2 Isolating Performance and Scalability Factors......Page 232
6.1.3 Applying Optimization and Tuning......Page 239
6.2 Effective Optimization and Tuning Techniques......Page 244
6.2.1 Wait Events and Service Demands......Page 245
6.2.2 Array Processing—Reducing V(i)......Page 247
6.2.3 Caching—Reducing Wait Time (W(i))......Page 250
6.2.4 Covering Index—Reducing Service Demand (D(i))......Page 252
6.2.5 Cursor-Sharing—Reducing Service Demand (D(i))......Page 253
6.2.6 Eliminating Extraneous Logic—Reducing Service Demand (D(i))......Page 255
6.2.7 Faster Storage—Reducing Data Latency (W(i))......Page 256
6.2.8 MPLS—Reducing Network Latency (W(i))......Page 257
6.2.9 Database Double Buffering—An Anti Performance and Scalability Pattern......Page 259
6.3 Balanced Queuing System......Page 264
6.4 Summary......Page 268
Recommended Reading......Page 269
Exercises......Page 270
PART 3 APPLYING API PROFILING......Page 273
7. Defining API Profiling Framework......Page 275
7.1 Defense Lines Against Software Performance and Scalability Defects......Page 276
7.2 Software Program Execution Stack......Page 277
7.3 The PerfBasic API Profiling Framework......Page 278
7.3.1 API Profile Logging Format......Page 279
7.3.2 Performance Log Parser......Page 280
7.3.3 Performance Maps......Page 282
7.4 Summary......Page 284
Exercises......Page 285
8. Enabling API Profiling Framework......Page 287
8.1 Overall Structure......Page 288
8.2 Global Parameters......Page 289
8.4 Processing Files......Page 290
8.5 Enabling Profiling......Page 291
8.6 Processing Inner Classes......Page 294
8.7 Processing Comments......Page 295
8.8 Processing Method Begin......Page 296
8.9 Processing Return Statements......Page 298
8.10 Processing Method End......Page 299
8.11 Processing Main Method......Page 300
8.12 Test Program......Page 301
Recommended Reading......Page 303
Exercises......Page 304
9.1 Graphics Tool—dot......Page 305
9.2 Graphics Tool—ILOG......Page 308
9.3 Graphics Resolution......Page 310
9.4.1 driver......Page 311
9.4.2 Global Parameters......Page 313
9.4.3 logReader......Page 315
9.4.4 logWriter......Page 316
9.4.6 Link......Page 317
9.4.8 utility......Page 318
9.4.9 parser......Page 319
9.4.10 xmlProcessor......Page 322
9.4.11 analyzer......Page 323
9.5 Summary......Page 324
Exercises......Page 325
10. Case Study: Applying API Profiling to Solving Software Performance and Scalability Challenges......Page 327
10.1 Enabling API Profiling......Page 328
10.1.1 Mechanism of Populating Log Entry......Page 329
10.1.3 Setting apf.properties File......Page 330
10.1.4 Parsing Workflow......Page 332
10.1.5 Verifying the Profiling-Enabled Source Code......Page 334
10.1.6 Recommended Best Coding Practices......Page 335
10.1.7 Enabling Non-Java Programs......Page 336
10.2.1 Generating API Profiling Log Data......Page 337
10.2.2 Parsing API Profiling Log Data......Page 338
10.2.3 Generating Performance Maps......Page 340
10.2.4 Making Sense Out of Performance Maps......Page 343
10.3.1 Using Adapter to Transform Custom Logs......Page 344
10.3.2 Generating Performance Maps with Custom Logs......Page 345
10.4.1 Client Side Performance Map......Page 349
10.4.2 Server Side Performance Map......Page 351
10.5.1 Baseline......Page 357
10.5.2 Optimization......Page 359
10.5.3 Analysis......Page 360
10.6 Summary......Page 361
Exercises......Page 362
A.1.1 Random Variables......Page 363
A.1.2 Random Variable Vector......Page 364
A.1.3 Independent and Identical Distributions (IID)......Page 365
A.1.5 Processes with Stationary Independent Increments......Page 366
A.2.3 Markov Processes......Page 367
A.3.1 Transition Probability Matrix and C-K Equations......Page 369
A.3.2 State Probability Matrix......Page 371
A.3.3 Classification of States and Chains......Page 372
A.4.2 Transition Rate Matrix......Page 373
A.4.3 Imbedded Markov Chains......Page 374
A.5.1 Definition......Page 375
A.5.2 Limiting State Probabilities......Page 377
A.5.4 Ergodic Theorems for Discrete-Time Markov Chains......Page 378
A.5.5 Ergodic Theorems for Continuous-Time Markov Chains......Page 380
A.6.1 Transition Rate Matrix......Page 381
A.6.2 C–K Equations......Page 382
A.6.4 Ergodicity......Page 383
APPENDIX B MEMORYLESS PROPERTY OF THE EXPONENTIAL DISTRIBUTION......Page 385
C.1 Review of Birth–Death Chains......Page 387
C.2 Utilization and Throughput......Page 388
C.4 Average System Time......Page 389
C.5 Average Wait Time......Page 390
INDEX......Page 391