Application of Agent- Oriented Techniques to Network Supervision Babak Esfandiari, Mitel Corporation.

Application of Agent-Oriented Techniques to Network Supervision

Babak Esfandiari,

Mitel Corporation

Different opportunities for agents in networks

Routing Network Management

Network Supervision GDMO/CMIS TMN

Network Supervision Problematic

Fault detection Alarm filtering and qualification Multiple and cascading faults ...

Existing attempts

Mostly use of expert systems for diagnosis ([Gaiti] [Garijo]…) Use of agent-oriented architectures (?) Revealed the importance of explicit

representation of Time No high-level communication between

network management platforms Acquisition of expertise is still a

bottleneck

ChroniclesChronicle RobotLoadMachine {

event (Robot: (outRoom, inRoom), e1);

event (Robot: (inRoom, outRoom), e4);

event (MachineInput: (unLoaded, loaded), e2);

event (Machine: (Stopped, Running), e1);

e1 < e2;

1’ <= e3 - e2 <= 6’;

3’ <= e4 - e2 <= 5’;

hold (Machine: Running, (e2, e2));

hold (SafetyConditions: True, (e1, e4));

when recognized {report “Successful load”;} }

Some theoretical speculations: Agents and OSI layers

Using Newell’s Knowledge Level as the highest communication layer? Expressing applications behaviors in

“rational” terms (Beliefs, Desires, Intentions, …)

Communicating such terms using high-level interaction languages (KIF/KQML?) and protocols

Interface Agents

“A program that […] provides assistance to a user dealing with a particular application. Such agents learn by watching over the shoulder of the user and detecting patterns and regularities…” (Maes)

The Assistant’s structure

eventaction CRS event

event

CLS action

action

ConfirmedChronicle

Base

UnconfirmedChronicle

Base

Network

Operator

Use of BDI to specify the agent’s behavior Trust as a modal operator B(a,f) Λ Trusts(a,b,f) -> K(a,f) Trusts(a,a,f) ? Trusts(a,human operator,*) Trusts(a,b,f) with b := other agent ?

Learnability of chronicles:a set of Oracles

PASSIVE: supplies events and actions PASSIVES: PASSIVE + no overlapping ACTIVEMQ: {events}+action -> yes/no

ACTIVEEQ:chronicle ->yes/(no+example)

Learnability of chronicles:Results

With one chronicle per action: positive with PASSIVES

positive with ACTIVEMQ+ PASSIVE

If more than one chronicle per action: negative with any oracle

Difficulties: overlapping x chronicles/action where find such oracles ?

The Learning System3 steps: Chronicle creation Chronicle evaluation Chronicle confirmation

An example (1)

Evaluation of: a b c -> Unconfirmed chronicle base:

1: a b c d -> Trust: 1

2: a b c e -> Trust: 1

3: a b c f -> Trust: 2

Confirmed chronicle base:

1: a b c g -> Trust: 3

An example (2)

Unconfirmed chronicle base:

1: d -> Trust: 1

2: e -> Trust: 1

3: a b c f -> Trust: 2

Confirmed chronicle base:

1: a b c g -> Trust: 3

2: a b c -> Trust: 3

MAGENTA: MAnaGEmeNT Application or Multi-AGENT Assistant ?

EM CM

MA

OM

CM EM

MA

OM

• ObjectManager: processes the query• CommunicationManager: sends and receives messages• EventManager: triggers event notifications• Management Application: processes the events and publishes queries

Experimentation

The local network Transpac data Robot behavior pattern detection Help to a Smalltalk programmer Overlapping management Collaborative learning

Results

Finding other oracles:Collaborative assistance

subnet subnet

operator operator

assistant assistant

• Presentation protocol• Matchmaking protocol• Query protocol

Conclusion and perspectives Summary:

Use of Interface Agents in Network Supervision Theoretical results on chronicle learning Appropriate algorithms

Use of Network Management standards to build an Agent Development platform

Next: Improve the algorithms (first order, partial order) Big scale experimentation Other applications of MAGENTA: remote

programming, distributed debugging, ...