ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 1

COST 285Modelling and Simulation Tools for Research in Emerging Multi-service Telecommunications

Francesco PotortìISTI (Istituto di Scienza e Tecnologie dell’Informazione)

CNR (Consiglio Nazionale delle Ricerche)Pisa – Italy

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 2

Main scientific interests

• Communication protocols, especially at the MAC level and transport level

• Satellite multiple access protocols in TDMA mode• Countermeasures against the atmospheric fading of

satellite signals• Mobile ad hoc networks with using 802.11 and Bluetooth

protocols, especially at the MAC level• Quality of service on the Internet• TCP over satellite and terrestrial wireless• Simulation in all of the above topics• Characterisation of wireless channels (802.11, GPRS)

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 3

Affiliation

• ISTI (Istituto di Scienza e Tecnologie dell’Informazione) is part of the CNR (Consiglio Nazionale delle Ricerche)

• It was born in Pisa, in 2002, as the result of a merger between CNUCE-CNR and IEI-CNR

• I am a member of the Wireless Networking Group, composed of about 10 persons

• I am currently involved into two research projects:– TANGO, an Italian research project aimed at multilayer,

multiservice Internet networks– IS-MANET, an Italian research project aimed at mobile ad hoc

networks in hostile environments

• I promote the use of free software in research environments

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 4

Simulation expertise

• Implementation of MTG, an Ethernet traffic generator used to measure the performance of FODA/IBEA

• Implementation and performance measurement of FODA/IBEA, a multiple-access satellite system

• Implementation of fracas, a discrete-time simulator for framed access channels, and its use to evaluate the performance of several satellite access systems

• Implementation of a fractal traffic generator• Implementation of GaliLEO, a prototypal event-driven

simulator for LEO satellite systems• Analytical analysis of TCP over satellite channels

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 5

Fracasthe Framed Channel Access Simulator

Francesco Potortì

• A very specialised simulator for communications protocols written in standard C

• Very small and very fast• Able to study the behaviour and the performance of

multiple-access protocols, usually at the MAC level• Used until now for satellite channel access protocols• Provides unsophisticated but comprehensive statistics• Wrapper in Python for independent replications

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 6

Scope of the simulation

• Multiple-access methods that use a time-framed communications channel

• In principle, usable for any such type of channel: the framed channel of FDDI2, but not FDDI, unless the time is artificially subdivided in frames, putting a lower boundary on time resolution

• In practice, always used for geostationary satellite access• The core is very small, consequently

– the access method can be as general as possible, with only the time frame constraint

– programming is not easy, and requires knowledge of all the inner structures of the simulator

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 7

General architecture

generator

generator

generator

generator

generator

generator

generator

Station 1

requester allocator

Station 2

Station 3

Station 4

allocation delay

Station1

Stat.2

Station3

Station4

frame allocations

statisticscollector

worker

worker

worker

worker

worker

worker

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 8

Traffic generators

• Each produces a number of Transmission Units per frame• Each generator is attached to a single station• Any number of generators can be attached to a station• Built-in generators include:

– two-state periodic fixed rate — can be used for one-shot– two-state periodic Poisson — can be used for one-shot– two-state Markov-modulated Poisson– two-dimensional (NxM states) Markov-modulated models VBR

traffic– fractional Gaussian white noise models generic aggregate traffic– specialised generators model batch interactive traffic

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 9

The stations and the allocator

• All traffic is expressed as the number of Transmission Units

• TRUs are produced by generators, queued at the stations, and sent according to the allocations in the current frame

• TRUs are not received; Fracas only simulates sending• For each frame, each station

– Queues traffic produced by attached generators– Drops traffic exceeding the queue length– Sends queued TRUs filling the allocation in the current frame– Asks the allocator for allocation in a future frame

• For each frame, the allocator sets up the allocation for the future frame as TRUs available to each station

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 10

Emulator core loop

do { frame_number += 1; for_all_stations_do {

input = run_generators (this_station); queue += input; sent = min (queue, allocation); queue -= sent; request = compute_request (this_station); } compute_allocations (frame_number + allocation_delay); gather_statistics (frame_number);} while (! Stop_condition ());run_workers_and_print_results ();

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 11

Allocator policies

• An allocator defines an allocation policy, which is an algorithm for computing the allocations for each station at a future frame based on their requests in the current frame

• Built-in policies are those that we really used in simulation work; others may be added as necessary

• Currently implemented policies:– fixed TDMA: a fixed assignment to each station– FODA/IBEA: developed at CNUCE, experimented with a prototype– VnL-DA: VBR allocation developed at CNUCE– FEEDERS: distributed allocation scheme developed at CNUCE– DRIFS : distributed allocation scheme developed at CNUCE– CFRA: developed at ENST - Toulouse (FR)

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 12

Output statistics

• At each frame, several observables are collected• Each worker computes a different statistics on one or more

observables, including:– TRUs input, queued, dropped, allocated, requested, sent by stations– allocation unused by stations– transmission delay, either per frame or per TRU

• Some workers produce their results at emulation time• Others produce a result at the end of the emulation• A Python wrapper around Fracas is used to obtain estimates

of a statistics inside a given confidence interval using independent replications of the same emulator run

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 13

Classes of traffic

• Fracas distinguishes among four different classes of traffic, hierarchically ordered as follows: stream, vbr, interactive, bulk, whose names reflect their supposed usage

• When emptying the station queues, the simulation core starts from stream, and gives any unused allocation to lower-grade traffic classes

• Usage of the classes is optional; possibilities include:– gathering statistics for traffic generated in different classes– defining different allocation policies for each class

• Use of classes enables the definition of complex allocation strategies

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 14

Fracas summaryFrancesco Potortì

• A lightweight, portable simulator specialised for the study of multiple-access allocation schemes

• Traffic generators, allocation policies and statistics computations can each be added as separate modules

• Many built-in modules are implemented and have been used in research work

• Fracas has been validated by checking against a prototypal implementation of the FODA/IBEA allocation algorithm

• A paper on the architecture of Fracas has been published on Telecommunications Systems in 1999

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 15

GaliLEO progress report

Laurent Franck and Francesco Potortì

• A comprehensive simulator for satellite constellations, targeted towards LEO/MEO communication systems

• Experience from previous projects (SimToc, LeoSim, Fracas), involves different academic institutions

• Free software entirely implemented in Java• Possible studies include:

– algorithms for cell frequency reuse– QoS routing for both UDL and ISL– QoS aware channel access techniques

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 16

• Emulation of an entire constellation

• Definition of individual earth stations and traffic generators

• Complete map of traffic patterns

• Ability to individually follow any traffic connection

Capacity of global coverage

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 17

• Emulation over a limited geographical region

• Detailed and realistic traffic generation– complex access techniques– complex frequency reuse

strategies

• Nearby network simulated by mathematical description

Study of a limited region

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 18

Basic terminology

• A UDL (up-down link) covers a satellite’s footprint

• A UDL is made of beams, each covering partially overlapping cells

• Satellites are connected by ISLs

• A station is fed by one or more traffic generators

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 19

Architecture of GaliLEO

• An event scheduler is at the core of the simulation engine

• The core modules come with the simulator

• Custom modules can be added at will

• Custom modules include traffic generators, constellation layouts, access protocols etc.

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 20

The two inner layers

• The simulation engine provides– the event scheduler– the framework for building modules– the communications facility between modules– the input file structure, statistics gathering and display facility

• The standard modules provide some basic features:– a Leonet and a polar constellation– a deterministic periodical traffic generator– scalar resources for allocation in stations and satellites– simple routing and load-dependent adaptive routing– basic allocation strategies for stations and satellites– basic station and satellite structures

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 21

The simulation engine

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 22

Demo setup

• 12 stations evenly distributed between 45°N and 45°S• Leonet constellation (15 satellites on 3 orbits)• ISL capacity is 20 connections, UDL’s is 80 connections• Measuring the connection blocking probability• For each connection, the first and last satellites are those

with max elevation• The demo is made of three steps, global traffic is the

same for all steps:– 6 Erlangs per station, static routing– 4 adjacent stations at 10 Erlangs, others at 4 Erlangs, static

routing– same traffic as previous case, adaptive routing

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 23

Demo results

ISTI - Consiglio Nazionale delle Ricerche – Pisa – ItalyISTI - Consiglio Nazionale delle Ricerche – Pisa – Italy 24

GaliLEO summaryLaurent Franck and Francesco Potortì

• A big simulator for LEO/MEO communication systems• Both overall earth view and geographically limited

studies are considered• Very modular; standard modules available, custom

modules written in Java can be added• Programming environment is Java• Work in progress: development is open to contributions• Some features usable today for research: demo available• Home page at http://galileo.tesa.prd.fr/