middleware solutions for simulation & modeling

24
Leila Jalali niversity of California, Irvine 2011 Spring SIW Interoperability of Multiple Autonomous Simulators in Integrated Simulation Environments Leila Jalali [email protected] http://www.ics.uci.edu/~ljalali/ Prof. Nalini Venkatasubramanian, Prof. Sharad Mehrotra University of California, Irvine

Upload: leila-jalali

Post on 09-May-2015

359 views

Category:

Technology


2 download

TRANSCRIPT

Page 1: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Interoperability of Multiple Autonomous Simulators in Integrated Simulation

Environments

Leila [email protected]

http://www.ics.uci.edu/~ljalali/

Prof. Nalini Venkatasubramanian, Prof. Sharad Mehrotra

University of California, Irvine

Page 2: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Simulation: the process of designing a model of a real world system and conducting experiments with this model for our purpose: cheaper, safer, easier, and quickerPlanning and decision support- defence simulations,

emergency response simulationsDomain specific Testing and Analysis - traffic analysis,

human behaviour study: crowd dynamics or evacuation simulators, network simulators

Immersive synthetic platforms for training

Introduction

Page 3: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Motivation for New Simulation Platforms

Many available simulators Operate on specific domains

e.g fire simulators, transportation simulators

Infeasible to build complex simulations entirely from scratchEconomic and organizational constraints The increasingly complex requirements

Need ability to:Bring together simulators from various modeling domains:

MetasimulationsModel and test larger and more complex scenarios Study cause- effect relationships to integrate simulators

Page 4: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Simulation Integration- historical view

1975 1980 1985 1990 1995 2000

SIMulator NETworking (SIMNET)(1983–1990) Combat Simulators

Distributed Interactive Simulation (DIS)(1990–today) Army Projects

Aggregate Level Simulation Protocol (ALSP)(1991–1997ish) War-gaming models

High Level Architecture (HLA) (1996-today) Defence

Defense Community

Adventure(Xerox PARC)

Dungeons and DragonsBoard Games Multi-User Dungeon (MUD)

Games

Internet & Gaming Community

Multi-User Video Games

Page 5: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Limitations of current approaches

Existing Integrated platforms, define a standard model and require the individual simulators to conform to the standardIt might not be always possibleThe standard may not have designed to handle the new

simulator needsCurrent model registration needs a lot of manual workThe approaches are costly, time consuming, easily fail,

difficult to maintain, difficult to scale

Most of other works on simulation integration provided specific services for interoperability in a small range of cases

HLA:─ Low level knowledge needed from the practitioner ─ Cost issues─ Complexity─ No support for semantic interoperability─ Transparency─ HLA is too big and mainly applied in defense

Page 6: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

General ChallengesManaging Complexity of Interoperating

SystemsAnalysis of cause- effect relationshipsReusability: e.g. components, modelsWe use meta models to describe simulator-

related meta-dataMake the underlying simulator more understandableAbstract of lower-level details of integration and

interoperability

CorrectnessEnsure the correctness of metasimulations

Time synchronization: timing issues and causality correctness

Data exchange: data transformations

Scalabilitye.g multiple geography

Page 7: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Reflective Architecture for Integrated Simulation Environments (RAISE)

INLET(Transportation Model)

Drillsim(Activity Model)

Fire, Earthquake(Crisis Model)

LTESim(Communication Model)S

IMU

LA

TO

RS

RA

ISE

M

idd

lew

are

Meta models

Analyzer & Adaptor Met

a le

vel

Complex Applications

External Data

Sources

Observe & Extract Reflect

Bas

e le

vel

Structural specification: UML diagrams, metamodelsInteractions: dependency sets, interdependent data

Lock Manager

Lock-table

Consistency Controller

dependencies

meta-actions

Synchronizer

Time SynchronizationData Exchange

Pub/Sub

OntologyTranslator

Page 8: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Using RAISE- step by step Reification

Extract simulators’ meta-data from base-level simulators (using the source code, interfaces, and databases) result in metamodels/specifications and data structures at the meta-level

Analysis of metamodels Extract the model elements and features that need to be integrated from metamodels Discover inter-dependencies

Run Federation Modified features of meta data structures that implement the integration are reflected

to the base-level simulators Ensuring the correctness

Time synchronization, Data management

Reification: Extract simulators’

meta-data

Analysis of metamodels:Discover inter-dependencies

Ensure Correctness:Time synchronizationData Transformations

Run Federation:Execute actions Communicate with metal-levelGenerate meta-actionsGenerate wrapper-actions

Pre-processing

Results Analysis

end of simulation?

no

yes

ParserDatabaseInterfacesSource code meta-data

meta-modelsinter-

dependencies

Page 9: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Major challenge: the complexity associated with reification

Creole as an Eclipse plug in Examine source code dependencies and to extract

the simulator’s features. Java simulators, not useful for complex and large

simulatorsA parser using a tool for large scale code

repositories searchExtract the entities and attributes from a

Java/Matlab simulator Simulator’s source codeInterfacesDatabases

Reification

Meta-level

Base-level

Reification Reflection

Page 10: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Making the underlying simulators more understandable

Abstracting out lower-level details of integration and interoperability

Need to be comprehensive and extensible

UML and Eclipse Modeling Framework

Metamodel

Base level

Meta level

Page 11: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Analyzer and Adaptor: to provide data transfer between simulators using data translators

Synchronizer: to monitor and control concurrent execution of multiple simulations• Using concepts from serializability theory in transaction processing • Developed three techniques: conservative, optimistic, hybrid

Prototype System Implementation

Page 12: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Synchronization in metasimulation

Ensuring causal correctness while preserving simulators’ autonomyA transaction-based approach to modeling the

synchronization problem by mapping it to a problem similar to multidatabase concurrency

A novel Hybrid Scheduling strategy for metasimulation synchronization which adapts itself to the "right" level of pessimism/optimism based on the state of the execution and underlying dependencies

Relaxation model (motivated by divergence control mechanisms and weak consistency models) which guarantee bounded violation of consistency

Applying proposed techniques in a detailed case study using multiple real-world simulators

Page 13: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Modeling Metasimulation

A metasimulation consists of a set of autonomous pre-existing simulators S1, S2 , S3 ,…, Sn that execute concurrently in an integrated environment

Using a transaction-based approach to modeling metasimulationsConsider each simulator’s execution as a sequence of

actions (time steps in time stepped simulators or events in event based simulators)

Scheduling multiple simulators actions such that dependencies be preserved

a three tuple Si=<Ti, Di , Ai> where:Ti : the type of the simulator

Time stepped or Event based Di : The data items that the simulator reads or updates. For

each data item, denotes the domain of d, which is a set of values that can be assigned to d.

Ai : the set of actions

Page 14: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Meta-synchronizer

Simulator i

. . .

Met

a le

vel

dependencies meta-actions

Metasimulation

d

Simulator j

d’

Bas

e le

vel

MetaSynchronizer

wrapper wrapper

. . .wrapperactions

wrapperactions

Meta-synchronizer: Upon receiving an external action from For all dependant simulators generate meta-action Post to meta-action queue Upon receiving a request Find all meta-actions from the queue s.t. and Send the metactions to Simulator’s wrapper: At the beginning of each iteration: t=current-time Send a request to get meta-actions Receive meta-actions Generate wrapper-actions At the end of each iteration: Send all external action that have been executed to meta-synchronizer

Page 15: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Address the synchronization problem by controlling the execution of the simulator's actions to ensure the legality of resulting schedules

Conservative Scheduling: ensures the legality of schedules by delaying the actions such that the dependencies are preserved in the concurrent execution of actions of different simulators

Optimistic Scheduling: we accept the fact that violations occur, resolve the violation when it does occur; by aborting the actions that caused the violation

Hybrid Scheduling: Combines the benefits of both the optimistic and conservative strategies (details not available in the slides)

Metascheduling strategies

Page 16: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Ideally, dependencies need to be reflected from one simulator into another as soon as update in one simulator becomes valid in another

In most of applications, ideal behavior results in unnecessary synchronization overhead and loss of concurrency among simulators.

Relax the dependencies that capture the extent to which simulators can deviate from ideal behaviorTime (t-bound): t-bound works as the delay

condition which states how much time the consumer can use a value behind the new update of the supplier

Value (v-distance): Let be the value of updated by and be the value of updated by , we consider the difference between the values of two data item using a user defined distance function

Number of changes (n-update): captures the maximum number of updates of supplier on before they become reflected on consumer

Relaxed Dependencies

Page 17: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

To validate the proposed reflective architectureUsing three disparate pre-existing simulators:

1. CFAST (Consolidated Model of Fire and Smoke Transport): a fire simulator

Simulates the effects of fire and smoke inside a building and Calculates the evolving distribution of smoke, fire gases and temperature

2. Drillsim: an activity simulator Multi-agent system that simulates human behavior in a

crisis

3. LTESim: a communication simulator Abstracts the physical layer and performs network level

simulations of 3GPP Long Term Evolution

A Case Study for simulation integration

Page 18: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Case study- simulators properties

Evacuation Simulator

Communication Simulator

Fire Simulator

DrillSim [9] Simulates a response activity evacuationTime steppedOpen source (in Java)Agent basedParameters: health profile, visual distance, speed of walking, num. of ongoing call, etc. Output: num. of evacuees, injuries, etc

LTESim [31]Performs network level simulations of 3GPP LTEEvent basedOpen source (in Matlab)Parameters: num. of transmit and receive antennas, uplink delay, network layout, channel model, bandwidth, frequency, receiver noise, etc. Output: pathloss, throughput, etc.

CFAST [10]Simulates the effects of fire and smoke inside a buildingTime steppedBlack-box (no access to source)Parameters: building geometry, materials of construction, fire properties, etc. Output: temperatures, pressure, gas concentrations: CO2, etc.

Page 19: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

CFAST Drillsim

Interaction between Fire simulation and Drillsim

smoke from fire can affect someone’s health

An Examlpe: CFAST - Drillsim Interaction

Agents Profile : HealthAgents Actions : Tell

People

Harmful conditions in each space at

any time

Page 20: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Metamodels

Page 21: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Inter-dependencies extracted from metamodels1. A harmful condition in CFAST can affect an individual’s

health in Drillsim.

2. Agents in Drillsim can communicate information on the fire and its location –increase the number of ongoing calls (people talk about the crisis) in Drillsim.

3. Harmful conditions in CFAST can affect the evacuation process in Drillsim, e.g. increase walking speed which maps to user speed in LTEsim.

4. Smoke in CFAST can decrease an agent’s visual distance in Drillsim.

5. The number of ongoing communications in Drillsim can affect network pathloss and throughput in LTEsim.

6. Pathloss in LTEsim can be used to determine connectivity/coverage in Drillsim.

7. Information on building layout from CFAST and Drillsim can determine the number of transmit and receive antenna required in LTEsim.

8. Number of evacuees from Drillsim determines the number of users in LTEsim.

9. Number of evacuees in Drillsim can affect receiver noise in LTEsim.

Page 22: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

(a) Average synchronization overhead in different simulation phases

(b)Total execution time in different simulation phases (c) Synchronization overhead vs. the number of

dependencies. (in (a) and (b) no. of dependencies=100)

Experiments

(a) (b) (c)

Page 23: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Hybrid Scheduling exhibits superior overall performance to other approaches

The choice of the approach is also dependent on the simulator, e.g. for event based simulators when the number of external events is large we need to avoid using OS

Relaxations always help into get better results in terms of synchronization overhead and total execution time

Experiments- conclusionStrategy CS CSR OS OSR HS HSRMetric synch. time synch. time synch. Time synch. time synch. time synch. time

CFAST 425.374 2225.626 348.812 2149.945 340.273 2140.273 309.931 2111.844 498.283 2298.475 316.007 2118.918

DrillSim 431.265 2232.235 331.192 2133.457 312.182 2113.165 252.011 2055.888 453.592 2253.698 288.555 2089.155

LTEsim 156.035 1956.530 99.277 1901.371 4887.753 3378.743 749.009 2550.043 344.005 2144.187 221.079 2023.039

Total 1012.674 6414.391 779.281 6188.723 2230.208 7632.181 1310.951 6717.755 1295.581 6696.360 816.641 6231.112

Page 24: Middleware Solutions for Simulation & Modeling

Leila JalaliUniversity of California, Irvine 2011 Spring SIW

Thanks

[email protected]://www.ics.uci.edu/~ljalali/