1 Simulation Methodology H Plan: –Introduce basics of simulation modeling –Define terminology and methods used –Introduce simulation paradigms u Time-driven simulation u Event-driven simulation u Monte Carlo simulation –Technical issues for simulations u Random number generation u Statistical inference

2 Performance Evaluation Analytical Methods Simulation Methods Experimental Methods

3 Performance Evaluation Analytical Methods Simulation Methods Experimental Methods Time-DrivenEvent-DrivenMonte Carlo...

4 Performance Evaluation Analytical Methods Simulation Methods Experimental Methods Time-DrivenEvent-DrivenMonte Carlo Sequential ParallelDistributed...

5 Time-Driven Simulation H Time advances in fixed size steps H Time step = smallest unit in model H Check each entity to see if state changes H Well-suited to continuous systems –e.g., river flow, factory floor automation H Granularity issue: –Too small: slow execution for model –Too large: miss important state changes

6 Event-Driven Simulation (1 of 2) H Discrete-event simulation (DES) H System is modeled as a set of entities that affect each other via events (msgs) H Each entity can have a set of states H Events happen at specific points in time (continous or discrete), and trigger state changes in the system H Very general technique, well-suited to modeling discrete systems (e.g, queues)

7 Event-Driven Simulation (2 of 2) H Typical implementation involves an event list, ordered by time H Process events in (non-decreasing) timestamp order, with seed event at t=0 H Each event can trigger 0 or more events –Zero: “dead end” event –One: “sustaining” event –More than one: “triggering” event H Simulation ends when event list is null, or desired time duration has elapsed

8 Sequential Simulation H Assumes a single processor system H Uses central event list (ordered by time) H Global state information available H Single, well-defined notion of time H Many clever implementation techniques and data structures for optimizing event list management –Linked list; doubly-linked list; priority queue; heap; calendar queue; trie structure

9 Parallel Simulation H Assumes multiple processors, often tightly coupled, with shared memory H Need fast inter-process communication H Shared state vs. no shared state H Event list: centralized or not? –Central event list can be a bottleneck –Decentralized requires careful coordination H Potentially different views of time H Conservative vs. Optimistic execution

10 Distributed Simulation H Assumes multiple processors, but geographically distributed (LAN/WAN) H Inter-process communication becomes expensive because of large latencies H Need to find right balance between computation and communication H Granularity of task scheduling H Similar technical issues to parallel simulation with respect to concurrency

11 Monte Carlo Simulation H Estimating an answer to some difficult problem using numerical approximation, based on random numbers H Examples: numerical integration, primality testing, WSN coverage H Suited to stochastic problems in which probabilistic answers are acceptable H Might be one-sided answers (e.g., prime) H Can bound probability to some epsilon

12 Summary H Simulation methods offer a range of general-purpose approaches for perf eval H Simulation modeler must determine the appropriate aspects of system to model H “The hardest part about simulation is deciding what not to model.” - M. Lavigne H Many technical issues: RNG, validation, statistical inference, efficiency H We will look at some examples soon