| 1 | package de.uka.ipd.sdq.reliability.solver.pcm2markov; |
| 2 | |
| 3 | import java.util.ArrayList; |
| 4 | import java.util.List; |
| 5 | import java.util.concurrent.TimeUnit; |
| 6 | |
| 7 | import org.apache.log4j.Logger; |
| 8 | |
| 9 | import de.uka.ipd.sdq.markov.MarkovChain; |
| 10 | import de.uka.ipd.sdq.pcm.usagemodel.UsageScenario; |
| 11 | import de.uka.ipd.sdq.pcmsolver.models.PCMInstance; |
| 12 | import de.uka.ipd.sdq.pcmsolver.runconfig.PCMSolverWorkflowRunConfiguration; |
| 13 | import de.uka.ipd.sdq.pcmsolver.visitors.UsageModelVisitor; |
| 14 | import de.uka.ipd.sdq.reliability.core.MarkovEvaluationType; |
| 15 | import de.uka.ipd.sdq.reliability.core.MarkovFailureType; |
| 16 | import de.uka.ipd.sdq.reliability.core.helper.MarkovFailureTypeHelper; |
| 17 | import de.uka.ipd.sdq.reliability.solver.MarkovSolver; |
| 18 | import de.uka.ipd.sdq.reliability.solver.sensitivity.MarkovSensitivity; |
| 19 | |
| 20 | /** |
| 21 | * This class has the functionality to perform a complete PCM2Markov |
| 22 | * transformation. |
| 23 | * |
| 24 | * @author brosch |
| 25 | * |
| 26 | */ |
| 27 | public class MarkovTransformation { |
| 28 | |
| 29 | /** |
| 30 | * A logger to give detailed information about the PCM instance |
| 31 | * transformation. |
| 32 | */ |
| 33 | private static Logger logger = Logger.getLogger(MarkovTransformation.class |
| 34 | .getName()); |
| 35 | |
| 36 | /** |
| 37 | * Provides functionality to manage failure types. |
| 38 | */ |
| 39 | private MarkovFailureTypeHelper helper = new MarkovFailureTypeHelper(); |
| 40 | |
| 41 | /** |
| 42 | * Checks whether there exists a next permutation based on the current one |
| 43 | * (which is given by the resources' current states). Generates the next |
| 44 | * permutation, if it exists, according to Narayana Pandita's method of |
| 45 | * systematically generating all permutations, and sets the resources' |
| 46 | * states accordingly. |
| 47 | * |
| 48 | * @param descriptors |
| 49 | * a list containing the processing resource descriptors |
| 50 | * @return true, if a new permutation was generated; false otherwise |
| 51 | */ |
| 52 | private boolean createNextPermutation( |
| 53 | final List<ProcessingResourceDescriptor> descriptors) { |
| 54 | int size = descriptors.size(); |
| 55 | |
| 56 | int k = -1; // used for step 1 (see below) |
| 57 | int l; // used for step 2 (see below) |
| 58 | MarkovResourceState tmp; // used for step 4 (see below) |
| 59 | int left; // used for step 4 (see below) |
| 60 | int right; // used for step 4 (see below) |
| 61 | |
| 62 | /* |
| 63 | * Systematic generation of all permutations according to Narayana |
| 64 | * Pandita. |
| 65 | */ |
| 66 | |
| 67 | // 1. Find the largest index k such that a[k] < a[k + 1]. If no such |
| 68 | // index exists the permutation |
| 69 | // is the last permutation. [wikipedia] |
| 70 | for (int i = 0; i < size - 1; i++) { |
| 71 | if (descriptors.get(i).getCurrentState() == MarkovResourceState.NA |
| 72 | && descriptors.get(i + 1).getCurrentState() == MarkovResourceState.OK) { |
| 73 | k = i; |
| 74 | } |
| 75 | } |
| 76 | if (k == -1) { |
| 77 | return false; // no such index exists (i.e., there is no next |
| 78 | // permutation) - return false |
| 79 | } |
| 80 | |
| 81 | // 2. Find the largest index l such that a[k] < a[l]. Since k + 1 is |
| 82 | // such an index, l is |
| 83 | // well defined and satisfies k < l. [wikipedia] |
| 84 | // Situation so far: a[k] = N/A, a[k+1] = OK. |
| 85 | // Now look for an l such that l > (k+1), and a[l] = OK. |
| 86 | l = k + 1; // a[k+1] is definitely OK. We may find another one, though: |
| 87 | for (int i = k + 2; i < size; i++) { |
| 88 | if (descriptors.get(i).getCurrentState() == MarkovResourceState.OK) { |
| 89 | l = i; // new such index found, update l |
| 90 | } |
| 91 | } |
| 92 | |
| 93 | // 3. Swap a[k] with a[l] [wikipedia] |
| 94 | // I.e., switch their states. |
| 95 | descriptors.get(k).switchState(); |
| 96 | descriptors.get(l).switchState(); |
| 97 | |
| 98 | // 4. Reverse the sequence from a[k + 1] up to and including the final |
| 99 | // element a[n]. [wikipedia] |
| 100 | for (left = k + 1, right = size - 1; left < right; left++, right--) { |
| 101 | // swap right and left |
| 102 | tmp = descriptors.get(left).getCurrentState(); |
| 103 | descriptors.get(left).setCurrentState( |
| 104 | descriptors.get(right).getCurrentState()); |
| 105 | descriptors.get(right).setCurrentState(tmp); |
| 106 | } |
| 107 | |
| 108 | return true; // next permutation successfully generated - return true |
| 109 | } |
| 110 | |
| 111 | /** |
| 112 | * Retrieves the failure types evaluation mode from a given analysis |
| 113 | * configuration. |
| 114 | * |
| 115 | * @param configuration |
| 116 | * the analysis configuration |
| 117 | * @return the evaluation mode |
| 118 | */ |
| 119 | private MarkovEvaluationType getEvaluationMode( |
| 120 | PCMSolverWorkflowRunConfiguration configuration) { |
| 121 | return MarkovEvaluationType.valueOf(configuration |
| 122 | .getMarkovEvaluationMode()); |
| 123 | } |
| 124 | |
| 125 | /** |
| 126 | * Sets the leftmost <i>m</i> resources in the descriptors list to N/A. |
| 127 | * |
| 128 | * @param descriptors |
| 129 | * a list containing the processing resource descriptors |
| 130 | * @param m |
| 131 | * the number of leftmost resources that will be set to N/A |
| 132 | */ |
| 133 | private void initializeResourceStates( |
| 134 | final List<ProcessingResourceDescriptor> descriptors, final int m) { |
| 135 | int size = descriptors.size(); |
| 136 | // set m leftmost resources to N/A |
| 137 | for (int i = 0; i < m; i++) { |
| 138 | descriptors.get(i).setCurrentState(MarkovResourceState.NA); |
| 139 | } |
| 140 | // set remaining (n-m) resources to OK |
| 141 | for (int i = m; i < size; i++) { |
| 142 | descriptors.get(i).setCurrentState(MarkovResourceState.OK); |
| 143 | } |
| 144 | } |
| 145 | |
| 146 | /** |
| 147 | * Checks whether a termination condition - possibly specified by the user - |
| 148 | * holds. |
| 149 | * |
| 150 | * @param configuration |
| 151 | * configuration properties for the reliability solver workflow |
| 152 | * @param markovResult |
| 153 | * results of the PCM2MarkovTransformation |
| 154 | * @param startTimeMs |
| 155 | * indicates when the process started. This needs to be a value |
| 156 | * in milliseconds determined by having called |
| 157 | * System.currentTimeMillis(). |
| 158 | * @return true, if a termination condition holds; false otherwise |
| 159 | */ |
| 160 | private boolean isStopConditionReached( |
| 161 | final PCMSolverWorkflowRunConfiguration configuration, |
| 162 | final MarkovTransformationResult markovResult, |
| 163 | final long startTimeMs) { |
| 164 | |
| 165 | // check solving time limit |
| 166 | if (configuration.isSolvingTimeLimitEnabled()) { |
| 167 | // solving time limit reached? |
| 168 | if (System.currentTimeMillis() - startTimeMs >= configuration |
| 169 | .getSolvingTimeLimit() * 1000) { |
| 170 | // yes, stop condition holds - return true |
| 171 | logger.info("Maximal solving time (" |
| 172 | + configuration.getSolvingTimeLimit() |
| 173 | + " seconds) reached - stopping!"); |
| 174 | return true; |
| 175 | } |
| 176 | } |
| 177 | |
| 178 | // check limit for the number of evaluated physical system states |
| 179 | if (configuration.isNumberOfEvaluatedSystemStatesEnabled()) { |
| 180 | // maximum number of system states to be evaluated reached? |
| 181 | if (markovResult.getPhysicalStateEvaluationCount() == configuration |
| 182 | .getNumberOfEvaluatedSystemStates()) { |
| 183 | // yes, stop condition holds - return true |
| 184 | logger.info("Maximal number of evaluated system states (" |
| 185 | + configuration.getNumberOfEvaluatedSystemStates() |
| 186 | + ") reached - stopping!"); |
| 187 | return true; |
| 188 | } |
| 189 | } |
| 190 | |
| 191 | // check limit for the number of exact decimal places |
| 192 | if (configuration.isNumberOfExactDecimalPlacesEnabled()) { |
| 193 | // required number of exact decimal places reached? |
| 194 | if (markovResult.hasRequiredAccuracy(configuration |
| 195 | .getNumberOfExactDecimalPlaces())) { |
| 196 | // yes, stop condition holds - return true |
| 197 | logger.info("Required number of exact decimal places (" |
| 198 | + configuration.getNumberOfExactDecimalPlaces() |
| 199 | + ") reached - stopping!"); |
| 200 | return true; |
| 201 | } |
| 202 | } |
| 203 | |
| 204 | // no stop condition holds - return false |
| 205 | return false; |
| 206 | } |
| 207 | |
| 208 | /** |
| 209 | * Solves all parametric dependencies within a given PCM instance. |
| 210 | * |
| 211 | * @param markovSource |
| 212 | * state information required during the PCM2MarkovTransformation |
| 213 | * @param scenario |
| 214 | * the usage scenario to evaluate |
| 215 | */ |
| 216 | private void runDSolver(final UsageScenario scenario, |
| 217 | final MarkovTransformationSource markovSource) { |
| 218 | |
| 219 | logger.debug("Resolving parametric dependencies."); |
| 220 | // Record the time consumed for solving parametric dependencies: |
| 221 | long startTime = System.nanoTime(); |
| 222 | |
| 223 | // The parametric dependencies are solved by using a visitor: |
| 224 | UsageModelVisitor visitor = new UsageModelVisitor(markovSource |
| 225 | .getModel()); |
| 226 | |
| 227 | // Solve the PCM instance using the visitor: |
| 228 | visitor.doSwitch(scenario.getScenarioBehaviour_UsageScenario()); |
| 229 | |
| 230 | // Let the user know about the time consumed: |
| 231 | long stopTime = System.nanoTime(); |
| 232 | long duration = TimeUnit.NANOSECONDS.toMillis(stopTime - startTime); |
| 233 | logger.info("Solved parametric dependencies: " + duration + " ms"); |
| 234 | } |
| 235 | |
| 236 | /** |
| 237 | * Transforms the given PCM instance into a Markov Chain instance. The PCM |
| 238 | * instance is assumed to have all parametric dependencies solved. |
| 239 | * |
| 240 | * @param configuration |
| 241 | * configuration properties for the reliability solver workflow |
| 242 | * @param scenario |
| 243 | * the UsageScenario to evaluate |
| 244 | * @param markovSource |
| 245 | * state information required during the PCM2MarkovTransformation |
| 246 | * @param markovResult |
| 247 | * results of the PCM2MarkovTransformation |
| 248 | * @return true if the transformation has stopped before completion, such |
| 249 | * that the result has to be approximated |
| 250 | */ |
| 251 | private boolean runPcm2Markov( |
| 252 | final PCMSolverWorkflowRunConfiguration configuration, |
| 253 | final UsageScenario scenario, |
| 254 | final MarkovTransformationSource markovSource, |
| 255 | final MarkovTransformationResult markovResult) { |
| 256 | |
| 257 | logger.debug("Transforming PCM model into analysis model."); |
| 258 | |
| 259 | // Declare the result variable: |
| 260 | boolean approximate = false; |
| 261 | |
| 262 | // Record the time consumed for creating the Markov Chain instance: |
| 263 | long startTime = System.nanoTime(); |
| 264 | |
| 265 | // Check for the requested state handling strategy: |
| 266 | if (!configuration.isIterationOverPhysicalSystemStatesEnabled()) { |
| 267 | // Run a single Markov transformation according to the |
| 268 | // "always ask" strategy: |
| 269 | runPcm2MarkovSingle(configuration, scenario, markovSource, |
| 270 | markovResult); |
| 271 | } else { |
| 272 | // Repeat the transformation for all physical system states: |
| 273 | approximate = runPcm2MarkovIteratively(configuration, scenario, |
| 274 | markovSource, markovResult); |
| 275 | } |
| 276 | |
| 277 | // Let the user know about the time consumed: |
| 278 | long stopTime = System.nanoTime(); |
| 279 | long duration = TimeUnit.NANOSECONDS.toMillis(stopTime - startTime); |
| 280 | logger.info("Finished Markov transformation: " + duration + " ms"); |
| 281 | if (configuration.isIterationOverPhysicalSystemStatesEnabled()) { |
| 282 | logger.info("Number of evaluated physical system states: " |
| 283 | + markovResult.getPhysicalStateEvaluationCount() |
| 284 | + " out of " |
| 285 | + markovResult.getNumberOfPhysicalSystemStates()); |
| 286 | } |
| 287 | |
| 288 | // Return the result: |
| 289 | return approximate; |
| 290 | } |
| 291 | |
| 292 | /** |
| 293 | * Runs the PCM 2 Markov transformation for all possible classes, or until |
| 294 | * at least one of the possible termination conditions holds. |
| 295 | * |
| 296 | * @param configuration |
| 297 | * configuration properties for the reliability solver workflow |
| 298 | * @param scenario |
| 299 | * the UsageScenario to evaluate |
| 300 | * @param markovSource |
| 301 | * state information required during the PCM2MarkovTransformation |
| 302 | * @param markovResult |
| 303 | * results of the PCM2MarkovTransformation |
| 304 | * @return true if the transformation has stopped before completion, such |
| 305 | * that the result has to be approximated |
| 306 | */ |
| 307 | private boolean runPcm2MarkovIteratively( |
| 308 | final PCMSolverWorkflowRunConfiguration configuration, |
| 309 | final UsageScenario scenario, |
| 310 | final MarkovTransformationSource markovSource, |
| 311 | final MarkovTransformationResult markovResult) { |
| 312 | |
| 313 | // Remember the start time of the transformation: |
| 314 | long startTimeMs = System.currentTimeMillis(); |
| 315 | |
| 316 | // Some short notations: |
| 317 | List<ProcessingResourceDescriptor> descriptors = markovSource |
| 318 | .getUnreliableResourceDescriptors(); |
| 319 | final int size = descriptors.size(); |
| 320 | |
| 321 | /* |
| 322 | * Now, for all n resources: Let m out of n (for m in 0..(n-1)) |
| 323 | * resources fail in an iteration, considering first the resources that |
| 324 | * are most likely to fail, moving on to resources that are less likely |
| 325 | * to fail, finally considering the resources that are least likely to |
| 326 | * fail. |
| 327 | */ |
| 328 | for (int permutationClass = 0; permutationClass <= size; permutationClass++) { |
| 329 | |
| 330 | /* |
| 331 | * Consider the current permutation first, where the m leftmost |
| 332 | * resources are set to N/A, while the remaining ones are OK. |
| 333 | */ |
| 334 | |
| 335 | // set m leftmost out of n resources total to N/A state |
| 336 | initializeResourceStates(descriptors, permutationClass); |
| 337 | |
| 338 | // markovSource.printCurrentResourceStates(); |
| 339 | |
| 340 | // Evaluate the physical system state: |
| 341 | runPcm2MarkovSingle(configuration, scenario, markovSource, |
| 342 | markovResult); |
| 343 | |
| 344 | // Check if any of the stop conditions is reached: |
| 345 | if (isStopConditionReached(configuration, markovResult, startTimeMs)) { |
| 346 | return true; |
| 347 | } |
| 348 | |
| 349 | /* |
| 350 | * Now generate all other permutations where m out of n |
| 351 | * ("n choose m") resources are N/A. |
| 352 | */ |
| 353 | for (;;) { |
| 354 | |
| 355 | // see if there's a next permutation |
| 356 | if (!createNextPermutation(descriptors)) { |
| 357 | break; // there's no new permutation - leave inner for loop |
| 358 | } |
| 359 | |
| 360 | // markovSource.printCurrentResourceStates(); |
| 361 | |
| 362 | /* |
| 363 | * At this point, we have a new permutation, i.e. system state. |
| 364 | * It was generated via the createNextPermutation(descriptors) |
| 365 | * call above. |
| 366 | */ |
| 367 | |
| 368 | // Evaluate the physical system state: |
| 369 | runPcm2MarkovSingle(configuration, scenario, markovSource, |
| 370 | markovResult); |
| 371 | |
| 372 | // Check if any of the stop conditions is reached: |
| 373 | if (isStopConditionReached(configuration, markovResult, |
| 374 | startTimeMs)) { |
| 375 | return true; |
| 376 | } |
| 377 | } // end for: generation of all "n choose m" permutations |
| 378 | } // end for: m out of n resources N/A |
| 379 | |
| 380 | // The whole transformation has been performed without reaching any stop |
| 381 | // conditions: |
| 382 | return false; |
| 383 | } |
| 384 | |
| 385 | /** |
| 386 | * Evaluates a single physical system state. |
| 387 | * |
| 388 | * @param configuration |
| 389 | * configuration properties for the reliability solver workflow |
| 390 | * @param scenario |
| 391 | * the UsageScenario to evaluate |
| 392 | * @param markovSource |
| 393 | * state information required during the PCM2MarkovTransformation |
| 394 | * @param markovResult |
| 395 | * results of the PCM2MarkovTransformation |
| 396 | */ |
| 397 | private void runPcm2MarkovSingle( |
| 398 | final PCMSolverWorkflowRunConfiguration configuration, |
| 399 | final UsageScenario scenario, |
| 400 | final MarkovTransformationSource markovSource, |
| 401 | final MarkovTransformationResult markovResult) { |
| 402 | |
| 403 | // Retrieve the MarkovSolver singleton instance: |
| 404 | MarkovSolver solver = MarkovSolver.getSingletonInstance(); |
| 405 | |
| 406 | // If Markov statistics printing is switched on, do the first physical |
| 407 | // state evaluation without Markov chain reduction, and count the number |
| 408 | // of resulting Markov states: |
| 409 | boolean countStates = configuration.isPrintMarkovStatistics() |
| 410 | && (markovResult.getPhysicalStateEvaluationCount() == 0); |
| 411 | |
| 412 | // calculate current state probability |
| 413 | double physicalStateProbability = 1.0; |
| 414 | if (configuration.isIterationOverPhysicalSystemStatesEnabled()) { |
| 415 | for (ProcessingResourceDescriptor descriptor : markovSource |
| 416 | .getUnreliableResourceDescriptors()) { |
| 417 | physicalStateProbability *= descriptor |
| 418 | .getStateProbability(descriptor.getCurrentState()); |
| 419 | } |
| 420 | } |
| 421 | |
| 422 | // The Markov Chain instance is created by using a visitor: |
| 423 | // MarkovUsageModelVisitor visitor = new MarkovUsageModelVisitor( |
| 424 | // markovSource, !countStates); |
| 425 | MarkovUsageModelVisitor visitor = new MarkovUsageModelVisitor( |
| 426 | markovSource, getEvaluationMode(configuration), !configuration |
| 427 | .isIterationOverPhysicalSystemStatesEnabled(), |
| 428 | configuration.isMarkovModelReductionEnabled(), configuration |
| 429 | .isMarkovModelTracesEnabled()); |
| 430 | |
| 431 | // Create the Markov Chain instance using the visitor: |
| 432 | MarkovChain resultChain = (MarkovChain) visitor.doSwitch(scenario |
| 433 | .getScenarioBehaviour_UsageScenario()); |
| 434 | |
| 435 | // Display information to the user: |
| 436 | if (countStates) { |
| 437 | logger |
| 438 | .info("Number of Markov states per evaluated physical system state:\t" |
| 439 | + resultChain.getStates().size()); |
| 440 | logger |
| 441 | .info("Number of Markov transitions per evaluated physical system state:\t" |
| 442 | + resultChain.getTransitions().size()); |
| 443 | } |
| 444 | |
| 445 | // Solve the Markov Chain, and add the result to the overall success |
| 446 | // probability, weighted by the probability of the current resource |
| 447 | // state combination: |
| 448 | double[][] markovProbabilityMatrix = solver.solve(resultChain); |
| 449 | |
| 450 | // Add the results of this evaluation: |
| 451 | markovResult.addPhysicalStateResults(resultChain, |
| 452 | markovProbabilityMatrix, physicalStateProbability); |
| 453 | } |
| 454 | |
| 455 | /** |
| 456 | * Transforms a PCM instance into a Markov Chain instance. |
| 457 | * |
| 458 | * The transformation is performed in two steps. In the first step, |
| 459 | * parametric dependencies within the PCM instance are solved using the |
| 460 | * dependency solver. The resulting PCM instance is then transformed into a |
| 461 | * Markov Chain. |
| 462 | * |
| 463 | * @param model |
| 464 | * the input PCM instance |
| 465 | * @param configuration |
| 466 | * configuration properties for the reliability solver workflow |
| 467 | * @param scenario |
| 468 | * the usage scenario to transform |
| 469 | * @return the transformation results |
| 470 | */ |
| 471 | private MarkovTransformationResult runScenarioTransform( |
| 472 | final PCMInstance model, |
| 473 | final PCMSolverWorkflowRunConfiguration configuration, |
| 474 | final UsageScenario scenario) { |
| 475 | |
| 476 | // Initialize failure type information: |
| 477 | List<MarkovFailureType> failureTypes = helper.getFailureTypes( |
| 478 | getEvaluationMode(configuration), model.getRepositories(), |
| 479 | model.getResourceEnvironment(), model.getSystem()); |
| 480 | |
| 481 | // Initialize state information: |
| 482 | MarkovTransformationSource markovSource = new MarkovTransformationSource( |
| 483 | model, true); |
| 484 | MarkovTransformationResult markovResult = new MarkovTransformationResult( |
| 485 | configuration, markovSource, scenario, failureTypes); |
| 486 | boolean approximate = false; |
| 487 | |
| 488 | // As a first step, solve parametric dependencies of the PCM instance: |
| 489 | try { |
| 490 | runDSolver(scenario, markovSource); |
| 491 | } catch (Exception e) { |
| 492 | |
| 493 | // The parametric dependencies could not be solved: |
| 494 | logger |
| 495 | .error("Solving of parametric dependencies caused exception: " |
| 496 | + e.getMessage() + " [" + e.getClass() + "]"); |
| 497 | e.printStackTrace(); |
| 498 | |
| 499 | return null; |
| 500 | } |
| 501 | |
| 502 | // Second, the PCM instance is transformed into a Markov Chain instance |
| 503 | // and solved for determining system reliability: |
| 504 | try { |
| 505 | approximate = runPcm2Markov(configuration, scenario, markovSource, |
| 506 | markovResult); |
| 507 | } catch (Exception e) { |
| 508 | logger.error("PCM 2 Markov transformation caused exception: " |
| 509 | + e.getMessage() + " [" + e.getClass() + "]"); |
| 510 | e.printStackTrace(); |
| 511 | } |
| 512 | |
| 513 | markovResult.setApproximate(approximate); |
| 514 | |
| 515 | // Return the transformation results: |
| 516 | return markovResult; |
| 517 | } |
| 518 | |
| 519 | /** |
| 520 | * Performs a PCM2Markov transformation. |
| 521 | * |
| 522 | * @param model |
| 523 | * the input PCM instance |
| 524 | * @param configuration |
| 525 | * configuration properties for the reliability solver workflow |
| 526 | * @param sensitivity |
| 527 | * sensitivity analysis configuration |
| 528 | * @return the transformation results |
| 529 | */ |
| 530 | public List<MarkovTransformationResult> runTransform( |
| 531 | final PCMInstance model, |
| 532 | final PCMSolverWorkflowRunConfiguration configuration, |
| 533 | final MarkovSensitivity sensitivity) { |
| 534 | if (sensitivity != null) { |
| 535 | return runTransformIteratively(model, configuration, sensitivity); |
| 536 | } else { |
| 537 | return runTransformSingle(model, configuration); |
| 538 | } |
| 539 | } |
| 540 | |
| 541 | /** |
| 542 | * Transforms a PCM instance into a Markov Chain instance. |
| 543 | * |
| 544 | * The transformation is repeated to enable sensitivity analysis. |
| 545 | * |
| 546 | * @param model |
| 547 | * the PCM instance |
| 548 | * @param configuration |
| 549 | * configuration properties for the reliability solver workflow |
| 550 | * @param sensitivity |
| 551 | * sensitivity analysis configuration |
| 552 | * @return the transformation results (only the results of the last step in |
| 553 | * the sensitivity analysis are returned) |
| 554 | */ |
| 555 | private List<MarkovTransformationResult> runTransformIteratively( |
| 556 | final PCMInstance model, |
| 557 | final PCMSolverWorkflowRunConfiguration configuration, |
| 558 | final MarkovSensitivity sensitivity) { |
| 559 | int sensitivityStepCount = 0; |
| 560 | List<MarkovTransformationResult> markovResults = null; |
| 561 | sensitivity.initialize(model); |
| 562 | PCMInstance step = sensitivity.getNextModel(); |
| 563 | while (step != null) { |
| 564 | sensitivityStepCount++; |
| 565 | logger.info("Starting sensitivity analysis step " |
| 566 | + sensitivityStepCount + "..."); |
| 567 | markovResults = runTransformSingle(step, configuration); |
| 568 | sensitivity.logResults(markovResults); |
| 569 | logger.info("Sensitivity analysis step " + sensitivityStepCount |
| 570 | + " completed"); |
| 571 | step = sensitivity.getNextModel(); |
| 572 | } |
| 573 | sensitivity.finalize(); |
| 574 | return markovResults; |
| 575 | } |
| 576 | |
| 577 | /** |
| 578 | * Transforms a PCM instance into a Markov Chain instance. |
| 579 | * |
| 580 | * The transformation is performed in two steps. In the first step, |
| 581 | * parametric dependencies within the PCM instance are solved using the |
| 582 | * dependency solver. The resulting PCM instance is then transformed into a |
| 583 | * Markov Chain. |
| 584 | * |
| 585 | * @param model |
| 586 | * the input PCM instance |
| 587 | * @param configuration |
| 588 | * configuration properties for the reliability solver workflow |
| 589 | * @return the transformation results |
| 590 | */ |
| 591 | private List<MarkovTransformationResult> runTransformSingle( |
| 592 | final PCMInstance model, |
| 593 | final PCMSolverWorkflowRunConfiguration configuration) { |
| 594 | |
| 595 | // Declare result list: |
| 596 | ArrayList<MarkovTransformationResult> resultList = new ArrayList<MarkovTransformationResult>(); |
| 597 | |
| 598 | // Transform all usage scenarios individually: |
| 599 | for (UsageScenario scenario : model.getUsageModel() |
| 600 | .getUsageScenario_UsageModel()) { |
| 601 | resultList |
| 602 | .add(runScenarioTransform(model, configuration, scenario)); |
| 603 | } |
| 604 | |
| 605 | // Return all transformation results: |
| 606 | return resultList; |
| 607 | } |
| 608 | } |