A Framework for Rulebased Dynamic Adaptation Ivan Lanese

A Framework for Rule-based Dynamic Adaptation Ivan Lanese Computer Science Department Focus research group Univers ity of Bologna/INRIA Bologna, Italy Joint work with Antonio Bucchiarone (Fondazione Bruno Kessler - IRST, Trento) and Fabrizio Montesi (University of Bologna/INRIA, Bologna) 1

Roadmap l Dynamic adaptation l A rule-based approach l The Jo. RBA prototype l Conclusions

Adaptation l Applications should change to adapt themselves to changes in the operating conditions – So to always meet user expectations – Without intervention from technicians l Adaptation can be triggered by – Dynamicity of the environment » Mobility » Communicating partners can appear and disappear – Changes in user needs –…

Built-in adaptation l l The application includes the logic to adapt itself Different approaches in the literature – Example: workflows extended with » Constraints on the environment » Handlers for environment events » … l Requires to foresee during application development most of the information concerning adaptation – Which event will trigger adaptation – Which new behavior will be required l This is not always possible

Dynamic adaptation l Aims at adapting applications to changes that were not predictable at application development time – Applicable also when built-in adaptation is not l Main challenges: – The adaptation logic cannot be part of the application – We want to update the application logic, not only the application state

Travelling scenario l l Bob is travelling from University of Bologna to University of Trento for a working meeting Bob has a mobile phone with a Travelling application – Booking the train ticket from Bologna to Trento – Instructing Bob to take the booked train – Booking the bus from Trento train station to Trento University – Instructing Bob to take the bus

High speed trains l l Assume a new high speed connection has been created between Bologna and Trento Applications involving taking trains may require adaptation – For instance, our Travelling application l When the Travelling application connects to Bologna train station information system, Bologna train station adaptation middleware – Discovers that the application may be adapted to exploit the new high speed connection – Checks if adaptation is possible » The application provides the needed adaptation interface – Checks if adaptation is desirable according to Bob’s preferences – Sends the updated code for booking the new high speed train

A few observations l l l The creation of an high speed connection was not expected when the travelling application has been developed (or even when it was started) Booking the high speed train may exploit a different protocol than booking other trains If the old connection still exists, Bob may still want to use it (e. g. , for price reasons)

How to implement this scenario? l Which kind of information about adaptation is required at application development time? – Difficult to adapt a black box (wrappers, …) l l How can such an adaptation be specified? How can those mechanisms be implemented in practice?

Roadmap l Dynamic adaptation l A rule-based approach l The Jo. RBA prototype l Conclusions

Our approach: main ideas l The application should provide an adaptation interface – A list of adaptable activities – And a few pieces of information about them l Adaptation is specified as a set of adaptation rules – One for each possible adaptation – External to the application – That can be created/modified after the application has been fully developed – Managed by an adaptation manager » Possibly implemented as a set of distributed adaptation servers

Our approach: an intuition

Adaptation interface l For each activity the adaptation interface should provide – A description of the activity » At least its name » Possibly its goal, a description according to some ontology, … – The set of variables that the activity uses to interact with the rest of the application – The non functional properties guaranteed by the current implementation – A comparator function allowing to compare two sets of non functional properties according to the user preferences

The Take Train activity l l Description: Take Train Variables: source (Bologna Train Station), destination (Trento Train Station), number (IC 2356) Non functional properties: time = 2 h 41 m, cost = 20€ Comparator function: new implementation is better if requires less time and each saved minute costs less than 0, 3€

Adaptation rules l An adaptation rule specifies when, where and how adaptation will be performed – Implements adaptation logic l Each rule includes: – A description of the activity to be adapted – An applicability condition (involving the state of the environment and the public state of the application) – The updated code for the activity – The set of variables needed by the new code to interact with the rest of the application – A state update – A compulsory flag – The non functional properties provided by the new activity

High speed train rule l l l l Description of the activity: Take Train Applicability condition: number = IC 2356 Updated code: code for booking and taking high speed train Variables: source, destination, number State update: number = FR 82 Compulsory flag: false Non functional properties: time = 1 h 23 m, cost = 32€

Triggering adaptation l l Adaptation can be either application-triggered or manager-triggered Application-triggered approaches – On initialization – On wait – On activity enter l Manager-triggered approaches – On registration – At time intervals – On rule update

Checking for adaptation l The adaptation manager checks the matching between a rule and an activity 1. 2. 3. 4. 5. The two descriptions should match All the variables needed should be available The applicability condition should be satisfied The compulsory flag should be true OR The non functional properties provided by the new implementation should be better than the old ones according to the user defined preferences

Applying adaptation l If all the conditions are satisfied adaptation can be applied – The new code is sent from the adaptation manager to the application replacing the old one – The non functional properties are updated – The state update is applied

Roadmap l Dynamic adaptation l A rule-based approach l The Jo. RBA prototype l Conclusions

Jo. RBA l l Jolie Rule-Based Adaptation framework A proof-of-concept implementation of our approach including – A skeleton for adaptable applications – An adaptation manager – A skeleton for adaptation servers l l Includes the implementation of the Travelling application Is written in Jolie

Jolie l l A language for programming service oriented applications Developed by Univerisity of Bologna spinoff Italiana. Software s. r. l. Implemented by a Java interpreter Takes inspiration from – sequential languages (assignment, if-then-else, while, …), – concurrent calculi (parallel composition, …) – workflow languages (one-way and request-response communication patterns) l Has native mechanisms for implementing loosely-coupled interactions such as the ones between adaptation servers and adaptable applications

Jo. RBA architecture l l l Jo. RBA exploits service oriented architectures to implement adaptation Adaptation managed by an Adaptation. Manager Rules managed by distributed Adaptation. Servers Applications can invoke the Adaptation. Manager to check for updates Activities are externalized as services, and updated thanks to dynamic embedding A state service provides support for the common state

Jo. RBA at work

Roadmap l Dynamic adaptation l A rule-based approach l The Jo. RBA prototype l Conclusions

Summary l l A novel approach for dynamic adaptation Based on the interaction between adaptable applications and external adaptation rules Very flexible framework A prototype implementation in Jolie

Future work l Still lot of work to do on the framework – – l l Descriptions of the activities Treatment of compulsory updates … (Anyone interested in working on the implementation? ) Applying the approach to other languages Lot of related topics to explore – Typing of rules and activities – Adapting ongoing activities – …

Finally