3.1 MB - Evernote

More documents

Recommendations

Info

91 Validation of the Parallel Simulator as part of the post simulation report found in the output logs of the simulate process for both simulation runs. For validation 5.1 the simulation performance was 1640 time units per second real time and for validation 5.2 1607 time units per second real time. The performance difference is quite small which suggests that for the example model used the processing saved on rolled back Transaction moves just about out weights the extra processing required for the LPCC, additional GVT calculations and the extra logging for the LPCC (the LP2 output log size of validation 5.2 is only around 3% of the one of validation 5.1). But for more complex simulation models where rollbacks in one LP lead to cascaded rollbacks in other LPs a much larger saving on the number of rolled back Transaction moves can be expected. It also needs to be considered that the hardware setup used (i.e. all nodes being run on a single CPU machine) is not ideal for a performance evaluation as the main purpose of this validation is to evaluate the functionality of the parallel simulator. 6.6 Validation 6 During the testing of the parallel simulator it became apparent that in same cases the normal Time Warp algorithm can outperform the Shock Resistant Time Warp algorithm. This last validation is showing this in an example. The simulation model used is very similar to the one used for validation 5. It contains two partitions with the first partition transferring some of its Transactions to the second partition but this time the GENERATE blocks are configured so that the first partition is ahead in simulation time compared to the second. The simulation is finished when 3000 Transactions have been terminated in one of the partitions. The complete simulation model can be seen here: PARTITION Partition1,3000 GENERATE 1,0,2000 TRANSFER 0.3,Label1 TERMINATE 1 PARTITION Partition2,3000 GENERATE 1,0 Label1 TERMINATE 1 Simulation model file model_validation6.gps
92 Validation of the Parallel Simulator As a result of the changed GENERATE block configuration and the first partition being ahead of the second partition in simulation time, all Transactions received by partition 2 from partition 1 are in the future for partition 2 and no rollbacks will be caused. But it will lead to an increase of the number of outstanding Transactions within partition 2 pushing up the number of uncommitted Transaction moves during the simulation. The logging configuration for this validation is also similar to the one used for validation 5 except that the LP statistic is not needed this time and is therefore switched off. The extract below shows the loggers for which debug output was enabled. … log4j.logger.parallelJavaGpssSimulator.lp=DEBUG log4j.logger.parallelJavaGpssSimulator.lp.commit=DEBUG log4j.logger.parallelJavaGpssSimulator.lp.rollback=DEBUG … log4j.logger.parallelJavaGpssSimulator.lp.lpcc=DEBUG log4j.logger.parallelJavaGpssSimulator.lp.lpcc.statespace=DEBUG log4j.logger.parallelJavaGpssSimulator.simulation=DEBUG … Extract of the used log4j configuration file proactive-log4j Like for validation 5 the script startNode.sh used to run the LP nodes is changed to extend the memory limit of the JVM to 128<strong>MB</strong>. Validation 6.1 For the first validation run the LPCC was enabled using the same configuration like for validation 5.1 resulting in the model being simulated using the Shock Resistant Time Warp algorithm. The significant effect of the simulation run is that the LPCC in LP2 starts setting actuator values in order to steer the local simulation processing towards a state that promises better performance but because the number of uncommitted Transaction moves within the second partition increases as a result of the Transactions received from partition 1 the actuator limits set by the LPCC are reached and the LP is switched into cancelback mode leading to its simulation progress being slowed down. In addition LP1 is also slowed down by the Transactions cancelled back from LP 2 as indicated by the output log of LP1. The actuator being set for LP2 can for instance be seen in line 25 to 28 of the output log of LP2. Subsequently the actuator limit is reached
Page 1 and 2:
TRANSACTION-ORIENTED SIMULATION IN
Page 3 and 4:
Table of Content II Table of Conten
Page 5 and 6:
IV Table of Content 6 VALIDATION OF
Page 7 and 8:
Figures VI Figures Figure 1: Ad Hoc
Page 9 and 10:
Tables VIII Tables Table 1: Change
Page 11 and 12:
2 Introduction synchronisation algo
Page 13 and 14:
2 Fundamental Concepts Fundamental
Page 15 and 16:
Figure 1: Ad Hoc Grid architecture
Page 17 and 18:
8 Fundamental Concepts In continuou
Page 19 and 20:
10 Fundamental Concepts Two other a
Page 21 and 22:
12 Fundamental Concepts known as st
Page 23 and 24:
14 Fundamental Concepts event execu
Page 25 and 26:
3 Ad Hoc Grid Aspects 16 Ad Hoc Gri
Page 27 and 28:
18 Ad Hoc Grid Aspects resources an
Page 29 and 30:
20 Ad Hoc Grid Aspects means that a
Page 31 and 32:
Parallel Transaction-oriented Simul
Page 33 and 34:
24 Parallel Transaction-oriented Si
Page 35 and 36:
4.2 Synchronisation algorithm 26 Pa
Page 37 and 38:
Page 39 and 40:
Page 41 and 42:
Page 43 and 44:
Page 45 and 46:
Page 47 and 48:
Page 49 and 50: 5 Implementation 40 Implementation
Page 51 and 52: 42 Implementation confirm a provisi
Page 53 and 54: 44 Implementation large numerical p
Page 55 and 56: 46 Implementation A detailed descri
Page 57 and 58: 48 Implementation But in a parallel
Page 59 and 60: 50 Implementation GPSS test models
Page 61 and 62: Simulation Controller side 52 Imple
Page 63 and 64: 54 Implementation technique describ
Page 65 and 66: Implementation Euclidean distance b
Page 67 and 68: 58 Implementation chain at the corr
Page 69 and 70: 60 Implementation The algorithm wil
Page 71 and 72: No Start Is Active Object body acti
Page 73 and 74: 5.3.2 GVT Calculation and End of Si
Page 75 and 76: 66 Implementation provisional simul
Page 77 and 78: Implementation relatively basic per
Page 79 and 80: 70 Implementation log4j.logger.para
Page 81 and 82: proactive-log4j 72 Implementation T
Page 83 and 84: 74 Implementation Memory Allocation
Page 85 and 86: 76 Validation of the Parallel Simul
Page 93 and 94: 6.4 Validation 4 84 Validation of t
Page 97 and 98: … LpccEnabled=true LpccClusterNum
Page 99: Validation 5.2 Validation of the Pa
Page 105 and 106: 7 Conclusions 96 Conclusions Even s
Page 107 and 108: References 98 References [1] Amin K
Page 109 and 110: [19] Krafft G. Parallelisation of T
Page 111 and 112: Appendix A: Detailed GPSS Syntax 10
Page 113 and 114: Reserved word: DEPART Syntax: [] DE
Page 115 and 116: Reserved word: SEIZE Syntax: [] SEI
Page 117 and 118: Setting: LpccClusterNumber Default
Page 119 and 120: Appendix C: Simulator log4j loggers
Page 121 and 122: Logger: parallelJavaGpssSimulator.l
Page 123 and 124: 114 Appendix D: Structure of the at
Page 125 and 126: Appendix E: Documentation of select
Page 131 and 132: !"#"$$%$&"'"(!))*+,-$"./#0$! 1$"))
Page 135 and 136: !/+/.7/8$ Appendix E: Documentation
Page 137 and 138: Returns: Appendix E: Documentation
Page 139 and 140: !"#$%&''$()*)")$ +',)$-)$. #,/. !"#
Page 141 and 142: !"#$%&'()#%*+,"-*.# Appendix E: Doc
Page 151 and 152:
Appendix E: Documentation of select
Page 153 and 154:
Constructor Detail !"#$%&'$(#$)*$%+
Page 155 and 156:
Appendix E: Documentation of select
Page 157 and 158:
Validation 1.1, output of simulate:
Page 159 and 160:
Validation 2, output of LP: Appendi
Page 161 and 162:
Appendix F: Validation output logs
Page 163 and 164:
Validation 2, output of simulate: A
Page 165 and 166:
Validation 3.1, output of LP1: Appe
Page 167 and 168:
Page 169 and 170:
Page 171 and 172:
Page 173 and 174:
Validation 4, output of LP1: Append
Page 175 and 176:
Page 177 and 178:
Page 179 and 180:
simulation for time 2000) (628) 20:
Page 181 and 182:
Page 183 and 184:
Page 185 and 186:
Page 187 and 188:
Validation 4, output of simulate: A
Page 189 and 190:
Page 191 and 192:
Page 193 and 194:
Page 195 and 196:
(135) 19:37:56,477 Total transactio
Page 197 and 198:
Page 199 and 200:
Page 201 and 202:
Page 203 and 204:
Page 205 and 206:
Page 207 and 208:
(68) 19:37:56,619 by the transactio
Page 209 and 210:
Page 211 and 212:
Page 213 and 214:
Page 215 and 216:
Page 217 and 218:
Page 219 and 220:
Page 221 and 222:
Page 223 and 224:
Page 225 and 226:
Page 227 and 228:
(1295) 14:25:22,740 Simulation stop
Page 229 and 230:
Page 231 and 232:
Page 233 and 234:
Page 235 and 236:
Page 237 and 238:
(68) 14:25:22,888 by the transactio
Page 239 and 240:
Validation 6.2, output of LP2: (1)
Page 241:
(32) (1,2) 0 765 Block: TRANSFER 0.
show all

3.1 MB - Evernote

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?