Presented By: Hector M Lugo-Cordero, MS EEL 6883 Programming Pervasive and Mobile Computing Applications: The TOTA Approach Presented By: Hector M Lugo-Cordero, MS EEL 6883
Full Citation Authors: Marco Mamei and Franco Zambonelli From: Università di Modena e Reggio Emilia Title: Programming pervasive and mobile computing applications: The TOTA approach Published at: ACM Transactions on Software Engineering and Methodology (TOSEM), vol. 18, issue 4, July 2009
Agenda Introduction Essentials Programming Evaluation Conclusions References
Agenda Introduction Essentials Programming Evaluation Conclusions References
The Problem Information processing is incorporated into everyday objects Agents can come and leave at any time Information is context (environment) dependent Need for a simple programming framework
TOTA A tupple oriented middleware Tuples are injected into the network No central common space
Case Study Scenario Imagine a huge museum like Le Louvre on France Many tourists come per day Makes hard for the management of services and information Assume every user has a wireless-enabled computer Museum layout can change over time and people come and go as they want
Current Approaches Direct communication Shared data-space Event base Communicate with other agents is direct (i.e. hardcoded) Not very dynamic Shared data-space Hard coordination Event base Notifications needed These are general (no context awareness)
Agenda Introduction Essentials Programming Evaluation Conclusions References
The TOTA Approach Gathers elements from both tuple-based and event-base models Distributed tuples injected to the network are cloned and propagated across the network A peer-to-peer network, each node running TOTA is the space with limited neighbors Tuples structure: T = (C, P, M) C = content, P = propagation, M = maintenance
TOTA Example (Information) With P = null C remains unchanged With M = null C is volatile
TOTA in Le Louvre Assumption that there is a large wireless network, with a backbone How to provide information and planning to avoid queues ArtPiece C = (description, location, distance) P = Propagate to all peers increasing distance M = update if topology changes
Solutions (Information) Art pieces propagate tuples Tourists query for the one with lower distance value Solution 2 Art pieces do not propagate tuples Users query for information Art pieces reply
TOTA Example (Meeting)
Solution (Meeting) Meeting C = (tourist_name, distance) P = propagate to all peers, increase distance by one M = update the distance tuple upon tourist move Tourist are guided with a GUI to the same place using the highest distance recursive process No coordination is specified by TOTA
TOTA Architecture
Agenda Introduction Essentials Programming Evaluation Conclusions References
TOTA Implementation Java based Using 802.11b broadcast Avoid unicast handshake Deployed emulator to increase network size Only 16 PDAs and some laptops were used The same code of the emulator could be used on devices Emulator ran in mixed modes (real and emulated devices)
TOTA Requirements What are the primitives that interact with the middleware? How to specify tuples T = (C, P, M)? How to code coordinated and context-aware activities?
TotaTuple public class MyTuple extends TotaTuple, implements ReactiveComponent{ //ReactiveComponent exposes the react() method } TotaTuple myTuple = new MyTuple(new Object[]{“Hello”});
Tota API
TOTA Predefined Tuples MessageTuple creates temporary tuples that expire when timeout occurs, ArtTuple HotTuple creates a tuple that reacts to topology changes by overloading the makeSubscriptions, MeetingTuple MetricTuple and SpaceTuple take into account physical location Metric updates each move based on a threshold Space only updates on the source
Programming ArtAgent1
Programming TouristAgent1
Programming ArtAgent2
Programming TouristAgent2
Programming MeetingAgent
Agenda Introduction Essentials Programming Evaluation Conclusions References
Software Engineering Analysis Context information makes TOTA general but more difficult for agents to react Separation of tuples (context) and agents (logic) however simplify this Incorporation of command and template design patterns facilitate tuple programming Command: Encapsulate a request as an object, thereby letting you parameterize clients with different requests, queue or log requests, and support undoable operations. Template: Define the skeleton of an algorithm in an operation, deferring some steps to subclasses. Template Method lets subclasses redefine certain steps of an algorithm without changing the algorithm's structure.
Tuple Propagation/Deletion Evaluation Highly scalable since only propagate to its immediate neighbors Tu = Trcv + Tprop + Tsend + Ttravel Propagation Time on a WiFi PDA (IPAQ 400 MHz) Tprop 99.7 ms Tsend 67.2 ms Ttravel 0 ms Trcv 21.2 ms Tu 188.1 ms
Tuple Propagation/Deletion Evaluation (cont.) For X hops TXU = X*TU In practice is a little more (10–20% from mean) Only HotTuples/MetricTuples and SpaceTuples were affected by topology
Load and Memory Evaluation Load is accounted to local agent execution and propagation/maintenance rules Storage is small enough for micro-sensors, but requires each one to store it Java implementation leaves TOTA opened for improvements in an optimized C version
Agenda Introduction Essentials Programming Evaluation Conclusions References
Conclusions TOTA support pervasive and mobile applications By using distributed tuples so agents can Extract contextual information Coordinate each other Dynamically adjusts as topology changes TOTA weaknesses Strict structure (hard mapping) Security and privacy issues Complex operations aren’t very supported
My Thoughts Strengths: TOTA seems like a powerful, portable API Weaknesses: Authors claimed that it is context independent but only the museum scenario was used Unclear mapping on how to transform real problems to TOTA tuples Suggestions: Adding flow of traffic by layer to see how to interpret the pool of tuples over the network Demonstrate the application on different domains, and give a clearer mapping Implementation in C for devices that do not posses a JVM Include wired backbone
Agenda Introduction Essentials Programming Evaluation Conclusions References
References http://www.wikipedia.org http://www.agentgroup.unimo.it/wiki/index.php/TOTA http://www.agentgroup.unimo.it/wiki/images/2/21/Tutorial.pdf http://www.agentgroup.unimo.it/wiki/images/8/8b/Tota.zip
Questions