Modern Software Engineering Methods for IP-QoS ... - TU Dresden

AQUILA Modern Software Engineering Methods for IP-QoS Resource Pool Management

Gerald Eichler T-Nova Deutsche Telekom Innovationsgesellschaft m.b.H. Technologiezentrum Darmstadt

Anne Thomas, Falk Fünfstück Dresden Technical University, Dept. of Computer Science, Institute SMT

Fabio Ricciato Consorzio di Ricerca sulle Telecomuniczioni - CoRiTeL, Roma

CoRiTeL

Charilaos Tsetsekas National Technical University of Athens Dept. of Electrical and Computer Engineering

Martin Winter Siemens AG, Munich, ICN WN CC EK A19 I2CS 2001 Ilmenau, June 21-22, 2001

1

Modern Software Engineering Methods for IP-QoS Resource Pool Management Gerald Eichler et. al. ([email protected])

AQUILA Outline ! Introduction: Gap between Applications and IP-Networks ! Resource Control Layer Model

• •

Architecture and Components Development Methodology

! Resource Offering

• •

Network Services and Application Profiles Accountability Knowledge Level & Item-Item Description Analysis Pattern

! Resource Management

• •

Resource Distribution Algorithms and Resource Pool Composite Structure Design Pattern

! Conclusion: Extension of the Basic Model

I2CS 2001 Ilmenau, June 21-22, 2001

2


AQUILA State of the Art ! Gap between Application Level and Network Management Level

• •

existing and upcoming QoS supporting models vs. classical traffic engineering methods

End-User’s View

Application Level

Resource Requests IntServ Model/RSVP

Traffic Prioritisation DiffServ Model

Resource Control Layer Resource Management Traffic Engineering

Resource Distribution Bandwidth Broker

Operator’s View I2CS 2001 Ilmenau, June 21-22, 2001

Network Management Level 3


AQUILA AQUILA’s Resource Control Layer ! AQUILA ... Adaptive Resource Control for QoS Using an IP-based Layered Architecture

• •

IST project of the Next Generation Networks Cluster (IST-1999-10077) URL: http://www-st.inf.tu-dresden.de/aquila/ Resource Control Layer

• IP/DiffServ Layer

SONET ATM

...

•

Physical Infrastructure Layer

introduction of additional Resource Control Layer agent based solution on top of a DiffServ aware IP-network Resource Control Agent IP/DiffServ aware router


4


AQUILA Resource Control Layer - Architecture Resource Control Layer Resource Control

ed

Us

Admission Control

QoS Request

s rce u o res

Admission Control Agent

QoS Request

Core Router Edge Router

Consideration of Network Load

Settings

QoS Request

Admission Control Agent

s rce u o res

Settings

End-user Application Toolkit

Resource Control Agent

Core Router

QoS Request

Monitoring Probing Results

End-user Application Toolkit

QoS Request

Edge Router

Core Router

Access Network

Access Network

IP Domain I2CS 2001 Ilmenau, June 21-22, 2001

5


AQUILA Resource Control Layer - Main Components ! Resource Control Agent (RCA)

• • •

keeps track of overall resource usage within an IP domain distributes resources to Admission Control Agents one instance per IP domain

! Admission Control Agent (ACA)

• • •

performs authentication, authorisation and admission control acts autonomously within assigned limits one instance per edge router

! End-user Application Toolkit (EAT)

• • •

offers Network Services to the customer translates and forwards QoS requests from the application to the ACA provides proxies for selected applications


6


AQUILA Software Engineering Process Evaluation of alternatives, new technologies & trends, e.g. MPLS

RCL Prototypes with new features

1

Planning for next cycle, brainstorming for new RCL features I2CS 2001 Ilmenau, June 21-22, 2001

2

3 4

8

1: 2: 3: 4:

RCL's Requirements Design of RCL components Development of RCL 1st trial ... 8: 2nd trial

06/2001 Intermediate & trial results, e.g. measurements 7


AQUILA AQUILA’s Software Development Support Overall Architecture

Design and Specification Task

Unified Modelling Language (UML) Analysis & Design Patterns Software Engineering I2CS 2001 Ilmenau, June 21-22, 2001

RCL Implementation

Module Communication

Interface Description

GUIs

Network Configuration Retrieval

Application related Data Exchange Implementation

Java Server Pages (JSP)

Java

Common Object Lightweight Directory Broker Architecture Access Protocol (CORBA) (LDAP) 8

Interface Definition Language (IDL)

eXtensible Markup Language (XML)

Technology


AQUILA Resource Offering - Network Services ! Objective

•

dynamical assignment of network resources to requesting applications

! Solution: Network Services

•

suited to data type (streaming multimedia, VoIP, ...) – Premium Constant Bit Rate, Premium Variable Bit Rate – Premium Multi-Media, Premium Mission Critical – Standard

• •

defined by the network operator composed of parameter sets – combination of QoS performance parameters ranges (delay, loss, ...) – set of traffic parameters ranges (peak rate, token bucket size, ...)

•

requested by application / end-user – high complexity of the request – mapping mechanism needed between end-user and network level


9


AQUILA Resource Offering - Mapping ! Application QoS Request Message at Network Level

• • •

list of low level network specific parameters include a Network Service Technical Characteristics

! Application QoS request at Application Level

• • •

correlated to the application type (VoIP, video conference, ...) verbal description of the wished quality (picture size, sound quality, ...) Session Characteristics

! Two way Mapping between both Levels done by the EAT

•

using the Application Profile information


10


AQUILA Resource Offering - Application Profile ! Application Profile (Definition)

• •

all relevant session and technical characteristics of an application described as data structure & analysis pattern

! Accountability Knowledge Level Analysis Pattern

•

operational level – corresponds to the concrete application

•

knowledge level – meta-level description (application type) – rules governing the structure

! Item-Item Description Analysis Pattern

•

relation between concrete objects and a shared description – each item object is described by a description item


11


AQUILA Analysis Patterns Knowledge level

Operational level

Accountability knowledge level Accountability type 0..*

Application

1..*

ApplicationType

e.g. video conference

e.g. NetMeeting Party

0..*

1..*

ServiceComponent 1

0..*

1..*

Party type

ServiceComponentType

1

Item

1..*

1

1

e.g. audio, video

1..* TechnicalCharacteristicType

Item-item description I2CS 2001 Ilmenau, June 21-22, 2001

Item description

1..*

SessionCharacteristicType 12


AQUILA Resource Management - Distribution ! Resource Distribution Algorithms

•

based on the “Hose Model”

•

Request Rule – Whenever the amount of used resources raises above the upper watermark, ask the root for a block of more resources per Network Service.

•

Release Rule – Whenever the amount of used resources falls above the lower watermark, return a block of free resources to the root per Network Service.

•

Dynamic Distribution – Dynamically shift resources within the group

•

use of the Resource Pool data structure


13


AQUILA Resource Management - Resource Pool ! Hierarchical Structure

• •

Domain

sub-area

“groups of groups” hierarchically organised tree

! Resource Limits

•

limit amount of QoS traffic from each edge router

subordinated sub-area

! Group neighboured Routers

•

limit amount of QoS traffic from each adjacent group

! Composite Structure Design Pattern

• •

data structure underlying the resource distribution algorithm representation in a tree structure (part-whole hierarchies)


14


AQUILA Composite Structure Design Pattern Component 1..*

ResourcePoolComponent

ResourcePool 1..*

Composite

interface ConsumableResourceShare

ResourcePoolLeaf

Leaf

1

1..*

Initialises 1

Manages 1

ResourceShare

ACA

1 Allocates / Deallocates

RCA

Composite I2CS 2001 Ilmenau, June 21-22, 2001

15


AQUILA AQUILA Project Goals for the 2nd Phase ! Scalability

•

single domain solution ⇒ multiple domain solution

! Flexible “Vertical” Resource Shift

•

dynamical resource sharing between Network Services

! Quality Adjustment

•

measurement driven control loops – to optimise admission control – to adapt router parameters (queuing, scheduling)

•

optimisation of traffic parameters for each Network Service

! Contribution to Standardisation

•

IETF drafts


16
