Home

User's Guide

1. 7 6 UniformStream arjuna JavaSim Distributions class UniformStream extends RandomStream public UniformStream double lo double hi public UniformStream double lo double hi int StreamSe public UniformStream double lo double hi int StreamSe long MGSeed long LCGSeed 9 lect lect lic double getNumber throws IOException ArithmeticException The JavaSim User s Manuals 7 7 ExponentialStream package arjuna JavaSim Distributions class Expone Te ExponentialS lic Exponentials ic ExponentialS blic double getNumber tream double m tream double m tream double m long MGSeed throws IOException ntialStream extends RandomStream int StreamSelect int StreamSelect long LCGSeed Arithmetic Exception 7 8 ErlangStream package arjuna JavaSim Distributions ic class ic lic ErlangStream ErlangStream lic ErlangStream lic double getNumber double mean double mean double mean long MGSeed throws do do do lo ErlangStream extends RandomStream uble sd uble sd uble sd ng LCGSeed Arithmetic int StreamSelect int StreamSelect
2. ecc a p Lb e Ure Abi pute MA Lp eo trem ie Hi 27 62 VARIANCB 5 u aa an sag 28 6 3 TIMEV ARIANGE c ener eei vcre eem eene eee ede 28 06 4 HISTOGRAMS 29 6 4 1 PrecisionHistOgram uso re iR e rer ret t eai Ur De sre ber ens 29 0 42 CHISIOBTAWEN a OUR RII E RR IR eeu e eee 30 64 3 SimpleHistogram can sive eee ii hene reb re d pti edere ae 31 64 4 Quanlle a aa nan ere ide Qa na 32 JAVASIM CLASSES E 33 TJ GSCHEDULER SD Aa T n D eae oak n a E a RID 33 7 2 34 13 SIMULATIONEXCEP TION a a ccuct Oa tee terne red 35 724 RESTARTEXCGEPTION kasqay 35 RANDOMSTREAM aha ie 35 7 6 UNIFORMSTREAM aaaasssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssssas 35 Tet EXPONENTIALSTREAM 36 7 8 TERLANGS TREAM qu a cna b
3. IOException void print DataInputStream iFile throws IOException The JavaSim User s Manuals 7 16 Variance package arjuna JavaSim Statistics ublic class Variance extends Mean ublic Variance public void setValue double value throws IllegalArgumentException public void reset ublic double variance p public double stdDev public double confidence double value public void print public boolean saveState String fileName throws IOException public boolean saveState DataOutputStream oFile throws IOException public boolean restoreState String fileName throws FileNotFoundException IOException public boolean restoreState DataInputStream iFile throws IOException 7 17 TimeVariance package arjuna JavaSim Statistics public class TimeVariance extends Variance public TimeVariance public void reset public void setValue double value throws IllegalArgumentException public boolean saveState String fileName throws IOException public boolean saveState DataOutputStream oFile throws IOException public boolean restoreState String fileName throws FileNotFoundException IOException eae double timeAverage public boolean restoreState DatalnputStream iFile throws IOException The JavaSim User s Manuals 7 18 Pr
4. public double confidence double value public void print public boolean saveState String fileName throws IOException public boolean saveState DataOutputStream oFile throws IOException public boolean restoreState String fileName throws FileNotFoundException IOException public boolean restoreState DatalnputStream iFile throws IOException variance returns the variance of the samples i 2 Si Mean Q m 1 n 1 pi II stdDev returns the standard deviation of the samples which is the square root of the variance 6 3 TimeVariance The TimeVariance class makes it possible to determine how long in terms of simulation time specific values were maintained In effect values are weighted according to the length of time that they were held whereas with the Variance class only the specific values are taken into account The JavaSim User s Manuals ublic class TimeVariance extends Variance ublic TimeVariance public void reset public void setValue double value throws IllegalArgumentException ublic boolean saveState String fileName throws IOException ublic boolean saveState DataOutputStream oFile throws IOException public boolean restoreState String fileName throws FileNotFoundException IOException public boolean restoreState DatalnputStream iFile throws IOException
5. PE SER I Eee PE ENTER e OE sessao 18 421 7 RANDOMSTREAM 18 42 UDNIFORMSTREAM ccccccccceeseecseseseseseseseseseseseseseseseseseseseeeseseseseseseseeeseseeeseseseseseceseseseceseseseseseseseeesesesees 19 4 3 EXPONENTIAESTREANMEQ Laqay 19 AA FRANGO R A 20 4 5 20 4 6 1 20 VIO aaa ae eT eee ee SE ee 21 48 S eI iS E A ETE 21 ADVANCED SIMULATION CLASSES 22 5 Ll 0 22 8 2 TRIGGER QUBUES 5 ete HR RO BER 24 3 9 GSEMAPHORES uang i nien 25 DAY Gegen Uie iU ede INNER 26 STATISTICAL CLASS ES ice ecco esee k awa epo sa eoa eda San repa ee ea T TS EVE ELS eva vo ee Q dE dew 27
6. double timeAverage Whenever a value is supplied to an instance of the TimeVariance class the simulation time at which it occurred is also noted If a value changes or the timeAverage method is invoked then the time it has been maintained for is calculated and the statistical data is updated 6 4 Histograms Mean Variance and TimeVariance provide a snapshot of values in the simulation However histograms can yield better information about how a range of values change over the course of a simulation run This information can be viewed in a number of ways but typically it is plotted in graphical form A histogram typically maintains a slot for each value or range of values given to it These slots are termed buckets and the way in which these buckets are maintained and manipulated gives rise to a variety of different histogram implementations The following sections detail this variety of different histogram classes 6 4 1 PrecisionHistogram The PrecisionHistogram class represents the core histogram class from which all others are derived This class keeps an exact tally of all values given to it 1 e a bucket is created for each value Although buckets are only created when requires over the course of a simulation this can still utilise a large amount of resources and so other less precise histogram classes are provided The JavaSim User s Manuals ublic class PrecisionHistogram extends Variance
7. execute the simulation Thread Type mainResume The final call to mainResume prevents run from exiting which we must do to ensure the application is portable between thread implementations 3 4 Resetting a simulation Resetting a simulation involves resetting all of the objects involved in it which will be required for subsequent runs When the reset method is invoked on the Scheduler this causes the current simulation run to be terminated and all simulation objects which are currently suspended on the scheduler queue will be woken and the RestartException will be thrown to each Any objects which are required to participate within a new simulation run must catch this exception reset themselves to a state consistent with the start of another simulation and then become suspended to await the restart of the simulation The JavaSim User s Manuals 3 41 Example If we take the Arrivals example above and add a reset method then the code could be public class Arrivals extends SimulationProcess public void run for try for Customer c new Customer queue insert Hold 20 0 catch RestartException e The JavaSim User s Manuals 4 Distribution Functions Many of the aspects of the real world which a simulation attempts to model have properties which correspond to various distribution functions e g inter arrival rates of customers at a bank queue Th
8. teDelay tivate lie tic tic gel Lace lc tig lic blic bilic c blic blic See SS ee SS ee oao Se ce blic blic lic lic void Re void Re void Re void Re void Re void Re void Re Ca te void void synchronized boolean idle boolean boolean static static static static ted Simu ted ted void void ted void ted void ted void ActivateBefore ActivateAfter ActivateAt ActivateAt ActivateDelay ActivateDelay Activate ncel rminate 0 passivated terminated SimulationProcess current double CurrentTim void mainSuspend void mainResume lationProcess set_evtime Hold double t Passivate Suspend Resume Exception RestartException ay lationException RestartException SimulationProcess throws Simul double AtTime double AtTime doub doub double time throws Simulation throws SimulationException SimulationProcess p throws Simul ationExcep Restar Exception tion p ationExcep Lion throws Simulation boolean prior Exception throws Simu RestartExcep tion la Restar e Delay throws Simul ationl tionExce
9. UNIVERSITY OF Object Oriented NEPA STE DION TUNE Discrete Event Simulation in Java JavaSim User s Guide Public Release 0 3 Version 1 0 Department of Computing Science Computing Laboratory The University Newcastle upon Tyne NE1 7RU UK The JavaSim User s Manuals Copyright Notice Copyright 1995 1999 Computing Laboratory University of Newcastle upon Tyne UK Permission to use copy modify and distribute the JavaSim software for evaluation teaching and or research purposes only and without fee is hereby granted providing that this copyright and permission notice appear on all copies and supporting documentation and that similar conditions are imposed on any individual or organisation to whom the program is distributed The University of Newcastle upon Tyne makes no representation about the suitability of this software for any purpose It is provided as is without express or implied warranty The JavaSim User s Manuals 1 2 3 Table of Contents PREFACE neenon uu Savon bossa de 0 a uu ua su aUe ec tes S u Sau ts 5 I1 vAWAIEABILTEY ate ete ir pl akuta EAA EE aie 5 INTRODUCTION P 6 2 SSIMUEATIONAMODBLS eU NINE 6 2 2 AEBRMINOEOGY rettet etre m retium He 6 2 3 CA
10. ublic PrecisionHistogram public void finalize public void setValue double value throws IllegalArgumentException public void reset public long numberOfBuckets IllegalArgumentException public double sizeByName double name throws IllegalArgumentException ublic boolean saveState String fileName throws IOException ublic boolean saveState DataOutputStream oFile throws IOException public boolean restoreState String fileName throws FileNotFoundException IOException public boolean restoreState DataInputStream iFile throws IOException pees double sizeByIndex long index throws StatisticsException public void print As with the Variance class from which it is derived and whose methods are obviously available values can be supplied to the histogram through the setValue double method The number of buckets maintained by the histogram can be obtained from the numberOfBuckets method Each bucket is uniquely named by the values it contains and can also be accessed by its index in the entire list of buckets There are therefore two ways of getting the number of entries in a bucket by the index number of the bucket sizeByIndex long index by the unique name of the bucket sizeByName double name If the bucket does not exist then each of these methods
11. Exception StatisticsException StatisticsException String s The JavaSim User s Manuals 8 References 1 2 8 G M Birtwistle O J Dahl B Myhrhaug Nygaard Simula Begin Academic Press 1973 O J Dahl B Myhrhaug K Nygaard SIMULA Common Base Language Norwegian Computing Centre Knuth Vol2 Seminumerical Algorithms Addison Wesley 1969 p 117 R Sedgewick Algorithms Addison Wesley Reading MA 1983 pp 36 38 D L McCue and M C Little Computing Replica Placement in Distributed Systems Proceedings of the 2 TEEE Workshop on the Management of Replicated Data November 1992 pp 58 61 I Mitrani Simulation Techniques for Discrete Event Systems Cambridge University Press Cambridge 1982 M C Little and D L McCue The Replica Management System a Scheme for Flexible and Dynamic Replication Proceedings of the 2 International Workshop on Configurable Distributed Systems March 1994 pp 46 57 G D Parrington et al The Design and Implementation of Arjuna Broadcast Project Technical Report October 1994 The JavaSim User s Manuals 9 Index Activating a process 12 Contact person 5 Controlling a simulation 14 examples ss S 14 Plattor os oes a Sasu sasa 5 DEIOGPSSES eiie across 10 class
12. an event list which indicates which events are to be scheduled for execution at specific simulation times Events are executed according to their simulation times There are two approaches to the way in which a simulator can schedule events to produce an operation path event oriented there is a procedure associated with each type of event in the system it performs the action required to handle that type of event and it is invoked every time such an event occurs In an event oriented approach an operation path is obtained by taking a global view of everything that happens in the system the manipulation of events is explicit e process oriented an operation path is obtained by the interacting of a number of processes running in parallel The management of events is implicit in the management of the processes The simulation system provides primitives for placing processes at particular points on the event list removing and re scheduling them The JavaSim User s Manuals The process oriented approach best fits with the object oriented paradigm which we want to present to the programmer of JavaSim As in SIMULA simulation processes then become active objects which interact with each other through message passing and the simulation primitives Refinements of these objects can then be obtained by inheriting from them and redefining the appropriate methods The JavaSim User s Manuals 3 Basic simulation classes This chapter describes
13. effectively monitoring the active and passive processes to enable it to determine when and which process to activate next A simulation application cannot affect the scheduler directly but can do so only indirectly through modifications of the scheduler queue Note the scheduler queue can be structured in a variety of ways including a linear list or a tree The implementation of the queue can depend upon the type of simulation being conducted For example a simulation which involves many concurrent processes would suffer from using a linear ordered queue which would typically have insertion and removal routines with overheads proportional to the number of entries in the queue However a linear 2In SIMULA the currently active process is not removed from the head of the queue The JavaSim User s Manuals list may work best for a low number of simulation processes JavaSim comes with a suite of scheduler queue implementations which can be chosen when the system is built 3 1 1 Scheduler class interface The scheduler is an instance of the Scheduler class It is the responsibility of the application programmer to ensure that only a single instance of this class is created public class Scheduler public static double CurrentTime public static synchronized void reset public static synchronized boolean simulationReset public static synchronized void startSimulation The scheduler maintains the si
14. either the CurrentTime or Time methods The former method is static and as such can be invoked without an instance of the SimulationProcess class Cancel removes the process from the scheduler queue or suspends it if it is the currently active process In either case the process is set to the passive state Passivate functions similarly but only works on the currently active process i e if it is invoked by the object e g through a publicly available method when it is not the current active process then it does nothing terminate removes the process from the scheduler queue or it is suspended if it is currently active The process is then set to the terminated state and can take no further part in this simulation run idle returns false if the process is either active or scheduled to become active Otherwise t rue is returned The JavaSim User s Manuals passivated and terminated indicate whether the process is in the passive or terminated state respectively Because SimulationProcess extends the java lang Thread class it is necessary for the simulation class to provide an implementation of the run method which will do the actual work for the process If this method ever returns then the thread is destroyed However in order for JavaSim to detect the termination of the thread the terminate method must be used instead 3 2 Example To illustrate how a simulation process could be implemented from the Pr
15. s Manuals 7 14 Semaphore package arjuna JavaSim Simulation public class SemaphoreOutcome 0 public int NOTDONE 1 public int WOULD_BLOCK static final static final static final public int DONE Pra class Semaphore public Semaphore public Semaphore long number public synchronized long NumberWaiting public synchronized int Get SimulationEntity toWait throws RestartException public synchronized int TryGet SimulationEntity toWait throws RestartException public synchronized int Release 7 15 arjuna JavaSim Statistics ublic class Mean ublic Mean lic void setValue double value 9 e throws IllegalArgumentException lic void reset 9 e O S g tO O 1 p pu 1 1 pu bli bli bl bli bli bli bli ic ie Le dc LG lle tic tic int numberOfSamples 0 double double double double boolean boolean boolean min max sum mean 0 0 0 0 saveState String fileName throws Exception saveState restoreStat DataOutputStream oFile boolean restoreStat throws IO String fileName throws FileNo cFoundl Exception Exception
16. the core elements of the JavaSim simulation system It is not intended as a tutorial on Java and basic knowledge of the language is assumed Note all of the classes described in this section can be found in the arjuna JavaSim Simulation package 3 1 The simulation scheduler Chapter 2 described the event list and how simulation entities processes are executed according to their position on the event list i e with increasing simulation time In JavaSim as in SIMULA simulation processes are managed by a scheduler and are placed on a scheduler queue the event list Processes are executed in pseudo parallel 1 only one process executes at any instance of real time but many processes may execute concurrently at any instance of simulation time The simulation clock is only advanced when all processes have been executed for the current instance of simulation time Inactive processes are placed on to the scheduler queue and when the current active process yields control to the scheduler either because it has finished or been placed back onto the scheduler queue the scheduler removes the process at the head of the queue and re activates it When the scheduler queue is empty i e there are no further processes left to execute the scheduler terminates the simulation Scheduler Queue Scheduler Active Process Figure 1 Scheduler Process Interaction As Figure 1 shows the scheduler co ordinates the entire simulation run
17. to occur which are outside the scope of those described above These will be described in the next section WaitForTrigger TriggerQueue queue places the current process on the trigger queue _queue and passivates it As with the previous methods the return value indicates whether the process was interrupted or triggered In addition to trigger queues process can wait on semaphores allowing the creation of monitor regions for example WaitForSemaphore Semaphore sem causes the current process to attempt to exclusively acquire the semaphore If this is not possible then the process is suspended Currently a process which is waiting on a semaphore cannot be interrupted and is not placed into the waiting state As such when this method returns the semaphore has been acquired 5 2 Trigger queues Processes waiting for the same application controlled event can be grouped together into a TriggerQueue as described in the previous section When this event occurs the application can use one of the two trigger methods to activate the queue members This involves placing the process es onto the head of the scheduler queue public class TriggerQueue public TriggerQueue public void finalize public synchronized void triggerFirst boolean setTrigger throws NoSuchElementException public synchronized void triggerFirst throws NoSuchElementException public synchronized void triggerAll throws
18. Exception Exception 7 9 HyperExponentialStream arjuna JavaSim Distributions pu bli bli bli bl lic ic class HyperExponentialStream extends RandomStream ic HyperExponentialStream double mean double sd ic HyperExponentialStream double mean double sd int StreamSelect ic HyperExponentialStream double mean double sd int StreamSelect long MGSeed long LCGSeed double getNumber throws IOException ArithmeticException 7 10 NormalStream arjuna JavaSim Distributions bl bli bl bli ic class NormalStream extends ic NormalStream double mean ic NormalStream double mean ic NormalStream double mean long MGSeed lic double getNumber throws Ra do do do lo ndomSt ream uble sd uble sd uble sd ng LCGSeed Exception Arithmetic int StreamSelect int StreamSelect Exception The JavaSim User s Manuals 7 11 Draw package arjuna JavaSim Distributions public class Draw public Draw double p public Draw double p int StreamSelect public Draw double p int StreamSelect long MGSeed long LCGSeed public boolean getBoolean throws IOException 7 12 SimulationEntity package
19. Exception tion Restart Res throws RestartException 0 0 Sake throws Simulation Exception tException Exception bl bl ic ic protected protected protected static static SimulationProcess void set_evtime void Hold void Passivate void mainSuspend void mainResume throws Simulat double time throws ionException Simulation double t throws SimulationException throws RestartException Exception RestartException The JavaSim User s Manuals 7 3 SimulationException package arjuna JavaSim Simulation public class SimulationException extends Exception public SimulationException public SimulationException String s 7 4 RestartException package arjuna JavaSim Simulation public class RestartException extends Exception public RestartException public RestartException String s 7 5 RandomStream arjuna JavaSim Distributions abstract class RandomStream public abstract double getNumber throws IOException ArithmeticException public final double Error 0 protected RandomStream protected RandomStream long MGSeed long LCGSeed protected final double Uniform
20. IOException ublic boolean saveState DataOutputStream oFile throws IOException IOException IOException The JavaSim User s Manuals 7 20 SimpleHistogram package arjuna JavaSim Statistics pub ic class SimpleHistogram extends PrecisionHistogram publ publ ic EG lic lic lic lic lic lic Tuc LEG EEG SimpleHistogram double min double max long nbuckets SimpleHistogram double min double max double w void setValue double value throws IllegalArgumentException void reset double sizeByName double name throws IllegalArgumentException double Width void print boolean saveState String fileName throws IOException boolean saveState DataOutputStream oFile throws IOException boolean restoreState String fileName throws FileNotFoundException IOException boolean restoreState DataInputStream iFile throws IOException 7 21 Quantile package arjuna JavaSim Statistics publ publ publ publ publ 1 EC ric class Quantile extends PrecisionHistogram Quantile Quantile double q throws IllegalArgumentException double getValue double range void print 7 22 StatisticsException package arjuna JavaSim Statistics publ publ publ lic lic lic class StatisticsException extends
21. NoSuchElementException triggerAl1l triggers all of the members on the queue e triggerFirst boolean setTrigger triggers only the head of the queue If set Trigger is true the default behaviour then the trigger method of the SimulationEntity object is also invoked If the queue is not empty when it is garbage collected by the virtual machine then all remaining queue members will be triggered and placed back onto the scheduler queue The JavaSim User s Manuals 5 3 Semaphores Application code can be protected from simulation processes through semaphores which are instances of the Semaphore class public class SemaphoreOutcome static final public int DONE 0 static final public int NOTDONE 1 static final public int WOULD_BLOCK 2 public class Semaphore public Semaphore public Semaphore long number public synchronized long NumberWaiting public synchronized int Get SimulationEntity toWait throws RestartException public synchronized int TryGet SimulationEntity toWait throws RestartException public synchronized int Release A semaphore can be used to restrict the number of processes which can use shared resources The number of shared resources available must be presented to the Semaphore when it is created By default a Semaphore will assume that there is only a single resource in which case a semaphore is exclusively acquired
22. TEGORIES OF SIMULATION 5 44 0 7 2 4 etienne uu 7 BASIC SIMULATION CLASSES 9 3 1 THE SIMULATION SCHEDULER 5 al evt usia aii EA exe eee e eue eue ye eee eb uec Ley ve E Ve Y Hua 9 ILI Scheduler class interf ce i see I een Res etre 10 3 2 SIMULATION 55665 10 321 Sim lationProcess class interface os e er e Pn eee edt 11 325 2 Examples lere esee ib qom ba 14 3 3 STARTING ENDING AND CONTROLLING A 8 120 0 202 2 6 0 0 32 00 00208 14 3 31 Suspending the main thread iss aue utere beret 15 9 9 2 SEXING the applic tiori eA eee e eR Ie treten eder pude eet dette delet 15 3 3 3 Controlling the simulation sei ie let ette eie b earned 16 3 4 RESETTING A SIMULATION u a a na n au Dee EE 16 3 419 Examples o ao e RR 17 DISTRIBUTION FUNCTIONS 5 es u ea ro re oe Pl SE se e DAVE asas YE
23. amSelect long MGSeed long LCGSeed p public ExponentialStream double mean p p public double getNumber throws IOException ArithmeticException The JavaSim User s Manuals StreamSelect indicates the offset in the random number sequence to begin sampling and MGSeed and LCGSeed can be used to modify the seed values used by the RandomSt ream class 4 4 ErlangStream ErlangStream returns an erlang distribution with mean mean and standard deviation sd public class ErlangStream extends RandomStream public ErlangStream double mean double sd public ErlangStream double mean double sd int StreamSelect public ErlangStream double mean double sd int StreamSelect long MGSeed long LCGSeed public double getNumber throws IOException ArithmeticException StreamSelect indicates the offset in the random number sequence to begin sampling and MGSeed and LCGSeed can be used to modify the seed values used by the RandomSt ream class 4 5 HyperExponentialStream The HyperExponential class returns a hyper exponential distribution of random numbers with mean mean and standard deviation sd public class HyperExponentialStream extends RandomStream public HyperExponentialStream double mean double sd public HyperExponentialStream double mean double sd int StreamSelect public HyperExponentialStream double mean do
24. arjuna JavaSim Simulation public class SimulationEntity extends SimulationProcess public void Interrupt SimulationEntity toInterrupt boolean immediate throws SimulationException RestartException public final void trigger public void terminate protected SimulationEntity protected void Wait double waitTime throws SimulationException RestartException InterruptedException protected void WaitFor SimulationEntity controller boolean reAct throws SimulationException RestartException InterruptedException protected void WaitFor SimulationEntity controller throws SimulationException RestartException InterruptedException protected void WaitForTrigger TriggerQueue gt queue throws SimulationException RestartException InterruptedException protected void WaitForSemaphore Semaphore _sem throws RestartException 7 13 TriggerQueue package arjuna JavaSim Simulation public class TriggerQueue public TriggerQueue public void finalize public synchronized void triggerFirst boolean setTrigger throws NoSuchElementException ublic synchronized void triggerFirst throws NoSuchElementException public synchronized void triggerAll throws NoSuchElementException The JavaSim User
25. by a simulation process However it is possible to create a Semaphore with different resource counts A Semaphore can exist in one of two states available the semaphore is available to be acquired unavailable a process or number of processes currently has the semaphore If another process attempts to acquire the semaphore then it is automatically suspended until the semaphore is available i e until a resource has been freed To be able to manipulate semaphores a process must be derived from the SimulationEntity class To obtain the semaphore the Get SimulationEntity toWait method should be used where t oWait is the calling process If the semaphore is unavailable then the process referenced by toWait is suspended If the semaphore is successfully acquired then SemaphoreOutcome DONE is returned otherwise SemaphoreOutcome NOTDONE If the process wishes to attempt to acquire the semaphore but does not want to block in the situation where the semaphore is currently unavailable then it can use the TryGet method which takes the same parameter as Get However unlike Get TryGet will return SemaphoreOutcome WOULD BLOCK in the case where the caller would normally block if it had called Get i e the semaphore is currently in use If the semaphore is not being used TryGet will acquire it for the caller Errors will result in SemaphoreOutcome NOT DONE being returned When the semaphore is no longer re
26. definition 11 Re activating a process 12 Resetting a simulation 15 Restarting a simulation 15 9 class definition 10 Simulation models 6 continuous time 7 continuous time discrete event 7 discrete time eter reta Z Terminolosy 6
27. e operation path is completely determined by the sequence of event times which need not be evenly spaced and can be of arbitrary increments and by the discrete changes in the system state which take place at these times i e the interactions of the events In between consecutive event times the system state may vary continuously Although it is possible to model the passage of real time by suitable event timing this is not necessary for a discrete event system the simulation model can advance its own internal time directly from one discrete event to another taking any appropriate action to advance the state accordingly It is this latter category of simulation modelling that JavaSim supports Examples of discrete event simulations are most queuing problems entities e g customers in a bank arrive according to a given distribution and change the system state instantaneously e g the number of customers in the queue The operation paths for this system are step functions they jump up or down by one when a customer joins or leaves the queue 2 4 Event scheduling Given that a simulation consists of a series of interacting events the operation path a simulator can be defined as that program devoted to the generation of operation paths The simulator allows the creation of events and controls their interactions according to a set of rules using an internal clock to keep track of the passage of simulation time It maintains
28. e throws FileNotFoundException IOException public boolean restoreState DataInputStream iFile throws IOException public void print New values can be supplied to the instance of the Mean class using the setValue double method The number of samples which have been give be obtained from numberOfSamples The maximum and minimum of the samples supplied can be obtained from the max and methods respectively sum returns the summation of all of the samples mean returns the mean value An instance of Mean can be reset between samples using the reset method The JavaSim User s Manuals If the state of a Mean object is required to be saved between simulation runs then it can be made persistent by using either of the saveState methods The first instance saves the state to a file whereas the second can be used to save the state to an instance of the java io DataOutputStream class There are likewise two corresponding ways in which the state can be restored The print method simply prints to System out the current state of the object 6 2 Variance This class is derived from Mean and in addition to providing the above mentioned functionality also provides the following public class Variance extends Mean public Variance public void setValue double value throws IllegalArgumentException public void reset public double variance public double stdDev
29. ean getBoolean throws IOException StreamSelect indicates the offset in the random number sequence to begin sampling and MGSeed and LCGSeed can be used to modify the seed values used by the RandomSt ream class 4 8 Example The JavaSim User s Manuals 5 Advanced Simulation Classes Simulations formed by the interaction of objects derived from SimulationProcess can be considered causal synchronous in nature events occur at specific times and form a well defined order However it is sometimes necessary to simulate asynchronous real world events e g processor interrupts To do this requires finer grained control of the scheduling of simulation processes than it provided by the scheduler the scheduler simply activates according to simulation time whereas asynchronous events may have different activation rules e g activate when another process is terminated The SimulationEntity class and others to be described in the following sections gives this required level of control to the user extending the types of simulation which are possible with JavaSim Asynchronous simulation processes are derived from SimulationEntity but the implementation enables these asynchronous process to execute in the same simulation as SimulationProcess objects However because these processes are suspended and resumed outside of the control of the scheduler it is possible for deadlock situations to occur Therefore some care must be taken when
30. ecisionHistogram package arjuna JavaSim Statistics public class PrecisionHistogram extends Variance public PrecisionHistogram public void finalize public void setValue double value throws IllegalArgumentException public void reset public long numberOfBuckets IllegalArgumentException public double sizeByName double name throws IllegalArgumentException ublic boolean saveState String fileName throws IOException ublic boolean saveState DataOutputStream oFile throws IOException public boolean restoreState String fileName throws FileNotFoundException public boolean restoreState DatalnputStream iFile throws public void print ore double sizeByIndex long index throws StatisticsException IOException IOException 7 19 Histogram package arjuna JavaSim Statistics public class Histogram extends PrecisionHistogram public Histogram long maxIndex int mergeChoice public Histogram long maxIndex public void setValue double value throws IllegalArgument public boolean restoreState DataInputStream iFile throws public void print ee boolean restoreState String fileName throws FileNotFoundException Exception ublic boolean saveState String fileName throws
31. ed by the call to the static mainSuspend method The code for main would then become public static void main String args c await Controller c new Controller 3 3 22 Exiting the application In order to exit a simulation application the application can call System exit However if it is only necessary to resume the main thread then this can be accomplished by using the static mainResume method of the SimulationProcess class Once the main thread has been resumed it will continue to execute from the point it was suspended In the example above this would be from within the await method The thread which calls mainResume can then suspend or terminate itself depending upon the application requirements The JavaSim User s Manuals public void exit if resumeMainRequired SimulationProcess mainResume Suspend else System exit 0 3 3 3 Controlling the simulation The controller s body creates and activates the other simulation entities and the scheduler and controls the overall simulation e g resetting the system between consecutive runs public void run sc new Scheduler create and activate any other simulation entities sc gt Resume we must create a scheduler for the simulation to run print results Sc Suspend suspend scheduler suspend simulation entities
32. elay teBefore teAfter teAt teAt teDelay teDelay SimulationProcess current double CurrentTime SimulationProcess p double AtTime throws Simul ationl NoSuchE SimulationProcess p throws Simul ement Exception Exception a tionl Exception Restar throws Simul a Exception ionl Exception Res boolean prior throws SimulationExcep tion throws Simula 5 tion double AtTime Restar boolean prior double Del throws ay Simulationl Exception double Del ay throws throws SimulationExcept Simu tionExcep Excep ti Re lationEx on star RestartException tar ion Res SimulationProcess throws Simulationl p Exception SimulationProcess throws Simulationl p Exception double AtTime boolean prior throws Simulationl double AtTime double Delay throws Simulationl Exception throws Simulationl LE m Re Re Re cept xcep star star star RestartExcep boolean prior double Delay throws SimulationException Exception Re tion s Exception tart Exception tart throws SimulationExcep Exception ion tion Exception tException Exception Exception
33. erefore simulation studies require sources of random numbers Ideally these sources should produce an endless stream of such numbers but to do so either requires specialised hardware or the ability to store an infinite large table of such numbers generated in advance Without such aids which are either impractical or not generally available the alternative is to use numerical algorithms No deterministic algorithm can produce a sequence of numbers that would have all of the properties of a truly random sequence 3 However for all practical purposes it is only necessary that the numbers produced appear random i e pass certain statistical tests for randomness Although these generators produce pseudo random numbers we continue to call the random number generators The starting point for generating arbitrary distribution functions is to produce a standard uniform distribution As we shall see all other distributions can be produced based upon this Interested readers are referred to 6 for a more complete treatment of this topic All of the distribution functions in JavaSim rely upon inheritance to specialise the behaviour obtained from the uniform distribution class These classes can be found the arjuna JavaSim Distributions package 4 1 RandomStream The actual uniform distribution class is called RandomStream This returns a series of random numbers uniformly distributed between 0 and 1 We experimented with several rando
34. g in the simulation In keeping with the object oriented paradigm and to make development of process objects simpler classes inherit the process functionality from the appropriate base class Process This class defines all of the necessary operations for the simulation system to control the simulation entities within it and for them to interact with it and each other At any point in simulation time a process can be in one and only one of the following states e active the process has been removed from the head of the scheduler queue and its actions are being executed suspended the process is on the scheduler queue scheduled to become active at a specified simulation time passive the process is not on the scheduler queue Unless another process brings it back on to the queue it will not execute any further actions The JavaSim User s Manuals terminated the process is not on the scheduler queue and has no further actions to execute Once a process has been terminated it cannot be made to execute further in the same simulation run A process which is either active or suspended is said to be scheduled 3 2 1 SimulationProcess class interface The SimulationProcess class definition is shown below Before considering how to build an example class derived from SimulationProcess we shall discuss the methods which it provides Because the constructors are protected it is not possible to create an instance of
35. imulation time specified by At Time The default for this time is the current simulation time The prior parameter is used to determine whether this process should be inserted before or after any processes with the same simulation time which may already be present in the queue The default is false ActivateDelay double AtTime boolean prior the process is activated after a specified delay At Time The process is inserted into the queue with the new simulation time and the prior parameter is used to determine its ordering with respect to other processes in the queue with the same time The default is false There are correspondingly five ReActivate methods which work on either passive or scheduled processes These will not be described in detail as they have similar signatures to their Activate counterparts and work in the same way Hold double period schedules the currently active process for re activation after the simulated delay of period time If this is invoked by the object e g through a publicly available method when it is not the current active process then it does nothing evtime returns the time at which the process is scheduled for activation next ev returns a reference to the next process to be scheduled for execution If the queue is empty then nu11 is returned The static method current returns a reference to the currently active process The current simulation time can be obtained by using
36. ined via the range method The actual quantile value is provided by get Value method The JavaSim User s Manuals 7 JavaSim classes For convenience in this section we shall include the JavaSim classes which programmers will use when constructing simulations 7 1 Scheduler arjuna JavaSim Simulation Pra class Scheduler public static double CurrentTime public static synchronized void reset public static synchronized boolean simulationReset public static synchronized void startSimulation The JavaSim User s Manuals 7 2 SimulationProcess arjuna JavaSim Simulation blic class SimulationProcess extends Thread T final double Time lic Tae synchronized SimulationProcess next_ev final double evtime lic void Activa void Activa lic void ActivateAt due void ActivateAt LG void Activa lae void Activa lic void Activate lic lic lic Tic lic lic lic blic blic blic blic blic blic void void void void void void void void void synchronized boolean idle boolean passivated boolean terminated static static ReActiva ReActiva ReActiva ReActiva ReActiva ReActiva ReActivate Cancel terminate teBefore teAfter teDelay teD
37. ject 8 Therefore it shares many of the same requirements easy to learn and use the interface to the simulation library should be easy to understand correct abstraction existing Java programmers should not find the simulation paradigm in conflict with the programming paradigm presented by Java Simulation programmers used to other environments should find the transition to JavaSim straightforward flexible and extensible it should be relatively easy for anyone to add new functionality to the system such as new distribution functions efficiency the system should be efficient and produce efficient simulation runs Simulation packages which we have experience of tended to be extremely slow and consume large amounts of system resources These requirements were realised in the following design decisions the discrete event process based simulation facilities provided by SIMULA 1 2 and its simulation classes and libraries have a considerable experience and user community which have found them to be successful for a wide variety of simulations In later versions of the system additional simulation classes were added which provide extra functionality inheritance was to be used throughout the design to even a greater extent than is already provided in SIMULA This enables JavaSim to be more flexible and extensible allowing new functionality to be added without affecting the overall system structure For example our I O facilities rando
38. m number generators and probability distribution functions are entirely object oriented relying on inheritance to specialise their behaviour 1 1 Availability JavaSim has been tested with JDK 1 0 2 and 1 1 x on Solaris linux and Windows 95 NT 4 0 Technical questions about JavaSim can be sent to M C Little ncl ac uk Tt is not necessary for the reader to know anything about SIMULA programming language or its simulation classes but such knowledge would aid in the understanding of the concepts and classes presented within The JavaSim User s Manuals 2 Introduction This manual is not intended as a tutorial on the concepts of simulation in general but rather how to write simulations in the JavaSim system However in order to be able to do this certain key simulation concepts will be briefly described The interested reader is referred to 6 for detailed descriptions of these concepts and for further discussions on simulation modelling 2 1 Simulation models To model a system is to replace it by something which is simpler and or easier to study equivalent to the original in all important respects Therefore before constructing the actual simulation it is first necessary to abstract from the real system those components and their interactions that are considered important for the actual model Building a simulation system model involves making certain simplifying assumptions to aid in the actual implementatio
39. m number generators before settling on a shuffle of a multiplicative generator with a linear congruential generator which provides a reasonably uniform stream of pseudo random numbers public abstract class RandomStream public abstract double getNumber throws IOException ArithmeticException public final double Error 0 protected RandomStream protected RandomStream long MGSeed long LCGSeed protected final double Uniform The multiplicative generator uses the following algorithm Y i 1 Y i 5 mod 226 where the period is get and the initial seed must be odd 3Thanks to Professor I Mitrani for his help in developing this The JavaSim User s Manuals The Uniform method uses the linear congruential generator seed is LCGSeed with the default value of 1878892440L based on the algorithm in 4 and the results of this are shuffled with the multiplicative generator see is MGSeed with a default value of 772531L as suggested by Maclaren and Marsaglia 3 to obtain a sufficiently uniform random distribution which is then returned The Error method returns a chi square error measure on the uniform distribution function By abstract method get Number must be provided by derived classes and is used to obtain a uniform means of accessing random numbers The RandomSt ream class returns a large sequence of random numbers whose period is 2424 However
40. mulation clock and the current value of this clock is obtained by invoking the CurrentTime method To enable multiple simulation runs to occur within a single application it is possible to reset the scheduler and simulation clock by calling the reset method This causes scheduler to remove all processes simulation objects currently registered on the scheduler queue and to invoke a class specific method on each of them which resets their states detailed in the next section Once this is finished the simulation is ready for an additional run A suspended process is informed that it has been reset by having the method it called to originally suspend itself ie place itself on the scheduler queue raise RestartSimulation exception which the object should catch It must then perform any work necessary to put itself back in a state ready for restarting the simulation and should then suspend itself again before the simulation can be restarted typically by calling Cancel 3 2 Simulation processes As was described in the previous chapter JavaSim supports the process oriented approach to simulation where each simulation entity can be considered a separate process Therefore in JavaSim the entities within a simulation are represented by process objects These are Java objects which have an independent thread of control associated with them at creation time allowing them to convey the notion of activity necessary for participatin
41. n and study of the simulation without such simplifications the model would be as complex as the system it is meant to be simulating However the accuracy of the results obtained from the simulation depend upon how valid the initial assumptions are For example when considering the trajectory of a projectile through the atmosphere the friction due to the air molecules is usually ignored This assumption is valid only within certain boundaries if the size of the projectile is on the same scale as the air molecules and its speed is sufficiently small then friction plays a significant role in its movement Thus the first step towards building a simulation model of a system is to determine exactly what are the important features which are to be measured and what characteristics of the system have an affect on them Any boundary conditions for the simulation e g size of projectile should be considered at the same time in order to simplify this procedure Building a final model can often take several phases where results from the initial model are compared with those obtained from the real system to determine their accuracy Any discrepancies are taken into account by possibly adding new components to the simulation until within certain error boundaries the simulation results match those from the real system 2 2 Terminology The system components chosen for the simulation are termed simulation entities Associated with each entity in the simula
42. ocess class we shall consider the example of a queue of customers arriving at a bank For this example this involves three classes Customer instances of this class represent the customers in the queue Queue the instance of this class queue is the queue into which customers are places Arrivals this is the process which creates new customers for insertion in queue The implementations of the Customer and Queue classes are not important to this example The implementation of the Arrivals class could be class Arrivals public Process public Arrivals Arrivals void Body Arrivals Body fot Customer c new Customer queue insert c Hold 20 0 3 3 Starting ending and controlling a simulation When a SimulationProcess object is created in JavaSim it starts in the passive state and must be activated before it can take part in the simulation This is typically performed by the first process object to which control is transferred after the simulation is initially started When writing JavaSim applications it is typical for the main thread to create a single controller process which is responsible for co ordinating the entire simulation run This creates and activates all of the simulation entities and the scheduler and provides methods for suspending the main thread thus allowing the controller object to execute and e
43. on public final void trigger public void terminate protected SimulationEntity protected void Wait double waitTime throws SimulationException RestartException InterruptedException protected void WaitFor SimulationEntity controller boolean reAct throws SimulationException RestartException InterruptedException protected void WaitFor SimulationEntity controller throws SimulationException RestartException InterruptedException protected void WaitForTrigger TriggerQueue _queue throws SimulationException RestartException InterruptedException protected void WaitForSemaphore Semaphore sem throws RestartException Because SimulationEntity is derived from SimulationProcess all of the usual simulation methods are available and can be used in conjunction with those provided by the derived class Interrupt SimulationEntity toInterrupt boolean immediate interrupts the asynchronous process toInterrupt which must not be terminated and must be in the waiting state toInterrupt becomes the next active process i e it is moved to the head of the scheduler queue If immediate is t rue then the current process is suspended immediately it is scheduled for reactivation at the current simulation time Otherwise the current process continues to execute and can be suspended later in an ap
44. plication specific way Because it is now possible for one process to wait for another to terminate the terminate method must differ from that provided SimulationProcess Before the terminating process ends it moves the waiting process to the head of the scheduler queue and then calls SimulationProcess terminate Currently only a single process can wait on this termination condition but this may change in future versions Wait double t is similar to Hold double t with the exception that the process is moved into the waiting state as well as being placed on the scheduler queue It is therefore possible to interrupt this process before the wait period has elapsed t rue is returned if the process was interrupted otherwise false is returned WaitFor SimulationEntity controller boolean reAct suspends the current process until controller has terminated The process is placed in the waiting state If reAct is true then cont roller is moved to the head of the scheduler queue to become the next activate process otherwise the default behaviour the application will have to activate controller If the waiting process is interrupted then the method returns The JavaSim User s Manuals true otherwise false The controller and the current process must be different 1 e it is not possible for a process to wait for itself Trigger queues are lists maintained by the simulation system of process waiting for specific events
45. ption tException boolean prior Exception e Delay RestartExcep throws Simu tion lationException RestartException throws SimulationException throws RestartException 0 0 RestartException throws SimulationException 0 0 throws Simulationl Exception throws SimulationException Exception RestartException throws RestartException throws RestartException RestartException RestartException There are five ways to activate a currently passive process which results in it being brought to the correct position in the scheduler queue corresponding to its associated simulation time If this is the head of the queue then it will become the active process The JavaSim User s Manuals e Activate this activates the process at the current simulation time ActivateBefore SimulationProcess proc this positions the process in the scheduler queue before proc and gives it the same simulation time If proc is not present then a SimulationException will be thrown ActivateAfter SimulationProcess proc this positions the process in the scheduler queue after proc and gives it the same simulation time If proc is not present then a SimulationException will be thrown ActivateAt double AtTime boolean prior the process is inserted into the scheduler queue at the position corresponding to the s
46. quired Release should be called by the process which currently has it Successful release of the semaphore results in The JavaSim User s Manuals SemaphoreOutcome DONE being returned otherwise Semaphore NOTDONE is returned NumberWaiting returns the number of processes currently suspended waiting for the semaphore If the semaphore is garbage collected with processes waiting for it then an error message is displayed No further action is attempted on behalf of these waiting processes 5 4 Example The JavaSim User s Manuals 6 Statistical Classes The purpose of a simulation typically involves the gathering of relevant statistical information e g the average length of time spent in a queue JavaSim provides a number of different classes for gathering such information These classes be found in the arjuna JavaSim Statistics package 6 1 Mean This is the basic class from which others are derived gathering statistical information on the samples provided to it public class Mean public Mean public void setValue double value throws IllegalArgumentException public void reset public int numberOfSamples public double min public double max public double sum public double mean public boolean saveState String fileName throws IOException public boolean saveState DataOutputStream oFile throws IOException public boolean restoreState String fileNam
47. the SimulationProcess class i e classes must be derived from this Processes are threaded objects and typically each thread package schedules execution of threads according to a priority By default all processes in JavaSim are created with the same priority but this can be altered by calling setPriority method of java lang Thread Note however that priorities have no effect on a simulation run public class SimulationProcess extends Thread public final double Time public synchronized SimulationProcess next ev throws SimulationException NoSuchElementException The JavaSim User s Manuals final double evtime FIG void Ac tivateBefore T Ke void Ac tivateAfter lu gel void Ac tivateAt lc void Ac tivateAt lic void Ac tivateDelay doub lic void Ac SimulationProcess p double AtTime double AtTime e SimulationProcess p throws Simulationl Exception boolean prior throws Simulation throws Simulation RestartException Exception RestartException Exception throws Simula RestartException tionException Restar Delay tiva doub iC c void Ac e throws Simulationl tException boolean prior De throws Simu
48. throws IllegalArgumentException It is possible to output the contents of the histogram to standard output using the print method 6 4 2 Histogram The problem with the PrecisionHistogram class is that it can use up a lot of system resources especially over the course of a long simulation Histogram attempts to alleviate this by presenting a histogram which is less accurate but consumes less resources Instead of maintaining a bucket for each individual value it keeps a fixed number of buckets Initially each bucket will store separate values as in the PrecisionHistogram but when the number of required buckets would exceed the specified maximum number it merges pairs of buckets thus reducing their total The policy used when merging buckets it set on a per instance basis when created Current policies are ACCUMULATE create a new bucket with the same name as the largest of the two buckets and it has the sum of the two old bucket entries as its entry number The JavaSim User s Manuals pu pu pu pu pu pu bli bli bl bl bli Fig LE LE Le ic Tc lic Lic MEAN create a new bucket with the name as the mean of the two old buckets and it has the sum of the two old bucket entries as its entry number MAX create a new bucket with the name as the largest of the two buckets and it has the same number of entries MIN create a new bucket with the name as the smallest of the t
49. tion are zero or more attributes that describe the state of the entity and which may vary during the course of the simulation The interaction of entities and the changes they cause in the system state are termed events The collection of these component attributes at any given time t defines the system state at t In general the system state can take any of a variety of values and a given simulation run results in one realisation of a set of these values the operation path over the observation period The JavaSim User s Manuals 2 3 Categories of simulation models There are three categories of simulation model described by the way in which the system state changes as a function of time Continuous time is one whose state varies continuously with time such systems are usually described by sets of differential equations Discrete time the system is considered only at selected moments in time the observation points These moments are typically evenly spaced Some economics models are examples of this where economics data becomes available at fixed intervals Changes in state are noticed only at observation points By choosing a suitably small interval between observation points a continuous time simulation can be approximated by a discrete time simulation Continuous time discrete event the time parameter is conceptually continuous and the observation period is a real interval usually starting at zero for simplicity Th
50. uble sd int StreamSelect long MGSeed long LCGSeed public double getNumber throws IOException ArithmeticException StreamSelect indicates the offset in the random number sequence to begin sampling and MGSeed and LCGSeed can be used to modify the seed values used by the RandomSt ream class 4 6 NormalStream NormalStream returns a normal distribution of random numbers with mean mean and standard deviation sd operator uses the polar method due to Box Muller and Marsaglia 3 public class NormalStream extends RandomStream public NormalStream double mean double sd public NormalStream double mean double sd int StreamSelect public NormalStream double mean double sd int StreamSelect long MGSeed long LCGSeed public double getNumber throws IOException ArithmeticException The JavaSim User s Manuals StreamSelect indicates the offset in the random number sequence to begin sampling and MGSeed and LCGSeed can be used to modify the seed values used by the RandomSt ream class 4 7 The Draw Draw class is the exception to the inheritance rule instead using RandomSt ream through delegation for historical reasons This returns t rue with the probability prob and false otherwise lic class Draw lic Draw double p lic Draw double p int StreamSelect lic Draw double p int StreamSelect long MGSeed long LCGSeed lic bool
51. una tee ove nuin iere eere iere RII 36 The JavaSim User s Manuals 7 9 HYPEREXPONENTIALSTREAM cccccsessscecececsesssececececsessseaeceeeceeseeaesecececsessaaeeeceeecseeaaeaeceesesensnneaeeeeeens 36 7 10 qaa AQ Pese e uus 36 7 11 Ib E EM M 37 7 12 SIMULATIONENTITY aaa awata EXER 37 7 13 E ne a uu 37 7 14 SEMAPHORE Q yayan anan ml ce Laven rete aD Prim 38 7 15 MEBAN Iona E P 38 7 16 RA VEN CUIM s 39 7 17 m 39 7 18 PRECISIONHISTOGRAM 40 7 19 HISTOGRAM HER 40 7 20 SIMPEEHISTOGRAM Gre ce ve ET E SEE 41 7 21 QUANTILE 5 ted vie io E 41 7 22 STATISTICSEXGEPTION 41 8 5 etie Seians 42 INDEN Gc 43 The JavaSim User s Manuals 1 Preface JavaSim is a Java implementation of the original C SIM simulation toolkit which was developed as a direct consequence of research conducted within the scope of the Arjuna pro
52. unless the seeds are modified when each random distribution class is created the starting position in this sequence will always be the same i e the same sequence of numbers will be obtained To prevent this each class derived from RandomSt ream has an additional parameter for one of its constructors which indicates the offset in this sequence from which to begin sampling 4 22 UniformStream The UniformStream class inherits from RandomSt ream and returns random numbers uniformly distributed over a range specified when the instance is created public class UniformStream extends RandomStream public UniformStream double lo double hi public UniformStream double lo double hi int StreamSelect public UniformStream double lo double hi int StreamSelect long MGSeed long LCGSeed public double getNumber throws IOException ArithmeticException The range covers the interval specified by 1o and hi StreamSelect indicates the offset in the random number sequence to begin sampling and MGSeed and LCGSeed can be used to modify the seed values used by the RandomSt ream class 4 3 ExponentialStream The ExponentialStream class returns an exponentially distributed stream of random numbers with mean value specified by mean ublic class ExponentialStream extends RandomStream ublic ExponentialStream double mean int StreamSelect ublic ExponentialStream double mean int Stre
53. using these classes In addition to the active suspended passive and terminated states which a simulation process can be in asynchronous objects can also be in the following states e waiting the process is suspended waiting for a specific event to occur e g a process to be terminated The waiting process is not placed on the scheduler queue interrupted the process which was in the waiting state has been interrupted from this before the condition it was awaiting occurred The conditions on which a process can wait and can thus be interrupted from are time a process can attempt to wait for a specified period of simulation time process termination a process can wait for the termination of another SimultionEntity process before continuing execution semaphore critical regions of a simulation can be protected by semaphores where only a single Entity process can acquire the semaphore other processes are suspended until the semaphore is released user specific it is possible for other asynchronous conditions to occur which are not covered above The classes to described in this chapter can be found the arjuna JavaSim Simulation package The JavaSim User s Manuals 5 1 Asynchronous entities public class SimulationEntity extends SimulationProcess public void Interrupt SimulationEntity toInterrupt boolean immediate throws SimulationException RestartExcepti
54. wo old buckets and it has the same number of entries class Histogram extends PrecisionHistogram Histogram long maxIndex int mergeChoice Histogram long maxIndex void setValue double value throws IllegalArgumentException boolean saveState String fileName throws IOException boolean saveState DataOutputStream oFile throws IOException boolean restoreState String fileName throws FileNotFoundException IOException boolean restoreState DataInputStream iFile throws IOException void print When an instance of Histogram is created the maximum number of allowed buckets must be specified The merging algorithm can also be provided with the default being the MEAN policy 6 4 5 SimpleHistogram As with the Histogram class above SimpleHistogram keeps the number of assigned buckets to a minimum However it does this by pre creating the buckets when it is created i e the number of required buckets must be provided at the start A width is the assigned for each bucket and whenever a value if given to the histogram class it is placed into the bucket whose width it falls within bj 511 bl 1 tic lic lic lic lic lic lic lic ric class SimpleHistogram extends PrecisionHistogram SimpleHistogram double min double max long nbuckets SimpleHistogram double min double max double w void setValue double value thro
55. ws IllegalArgumentException void reset double sizeByName double name throws IllegalArgumentException double Width void print boolean saveState String fileName throws IOException boolean saveState DataOutputStream oFile throws IOException boolean restoreState String fileName throws FileNotFoundException IOException boolean restoreState DataInputStream iFile throws IOException When the class is instantiated the range of values it will receive must be provided Then either the width of each bucket or the actual number of buckets can be given If the width is The JavaSim User s Manuals provided then the histogram automatically calculates the number of buckets otherwise it calculates the width for each bucket by equally dividing the range between each bucket The values of a bucket can be obtained from the sizeByName method The width of each bucket is provided by the Width method 6 4 4 Quantile The Quantile class provides a means of obtaining the p quantile of a distribution of values i e the value below which p percent of the distribution lies public class Quantile extends PrecisionHistogram public Quantile public Quantile double q throws IllegalArgumentException public double getValue public double range public void print The p quantile probability range must be specified when the object is instantiated and can be obta
56. xiting the application An example controller interface is shown below and the implementations for its methods will be described in the following sections The JavaSim User s Manuals ublic class Controller extends SimulationProcess ublic Controller public void await public void exit public void run Because Controller is a simulation process itself it derives from Simulationrocess and defines a run method which will do the actual controlling of the simulation It also provides the following methods await this method is called within the main application thread and suspends it effectively transferring control the Controller process exit this method is called to exit the simulation 3 3 1 Suspending the main thread When a threaded application is started it is important to realise that before any application threads are created the Java virtual machine has already created one to run the application This thread must be suspended before any simulation threads can run The await method of Controller is responsible for suspending this thread public void await SimulationProcess mainSuspend Resume It must first resume the thread associated with the Controller instance since Controller isa SimulationProcess it starts in the passive state This thread does not execute until the main thread is suspend

User's Guide

Contents

Download Pdf Manuals

Related Search

Related Contents