A novel online fucntional testing methodology
Beginning : 2015 |
Action line : SciLex 1
Subject : Distributed spaces in concurrent in espistemic systems.
Directors : Stéphane MAAG, Fatiha ZAÏDI
Institution :SAMOVAR ; LRI
Administrator laboratory : SAMOVAR
Phd Student : Jose-Alfredo ALVAREZ-ALDANA
Defense thesis : september 28, 2018
Scientific productions :
Several computational models are used by different actors (e.g., ISP, network administrators, software managers, etc.) to monitor and test specific system properties or behaviors through diverse spread probes. However, these current computational testing models can overload the communication architecture, be inefficient or even harmful, in particular in distributed complex systems. In this context, several probes or observers need to be spread to provide altogether a service defined as a function. The main research objective of this thesis is to propose a novel formal passive testing approach for complex distributed systems based on continuous monitoring (by considering communication costs, optimized bounds, space complexities, etc). We aim an improvement of the testing verdicts of distributed systems by online continuous monitoring.
When surveying the current solutions for continuous monitoring distributed systems, we note that many of them still proposes centralized approaches or classical and periodic polling that lead to important communication cost and impact the online and continuous test verdicts 26. The main challenges we expect to tackle through this thesis are :
- To compute better bounds reducing the complexity space and time consumed by the observers/ probes.
- A smart fully distributed correlation of the observed data streams through novel “functional polling” techniques (e.g. redundant inputs) with online-tracking proposals.
- Online testing by continuous distributed monitoring.
- Integrating our solutions on a real industrial testing platform. Currently, there is no real deployment of continuous distributed monitoring approaches, most solutions and concepts being either theoretical or simulated.
The main scientific objectives of this thesis are the following :
- To study the current formal approaches to passively test and continuously monitor distributed systems.
- To propose a novel distributed methodology to test distributed functions in communication systems. Model-based functional monitoring will be studied and considered.
- To study and propose optimized bounds of communication cost in general. This goal in distributed systems is crucial when passively testing communicating entities.
- To apply the proposed solutions to systems that are fully distributed and data centric such as IP Multimedia Subsystems, cloud systems or wireless sensor networks. We also guess that our proposed methodology could tackle a gap between distributed testing and anomaly detection in networks.
- Finally, transfers through eventual industrial collaborations or patenting aspects are goaled.
The main results of this PhD Thesis are:
We provided a preliminary survey and abstract of the monitoring of mobile ad-hoc networks literature. We focused on the significant problematics which can be summarized to the dynamic nature of the networks since they are affected by multiple external parameters. We defined that for monitoring a network, it can be done with a centralized, decentralized or distributed approach. Although the centralized approach sounds more comfortable and efficient for wired networks, it is not the case when applied to a MANET. Therefore the decentralized and distributed approaches become an interesting and exciting alternative to achieve monitoring. We then propose a decentralized approach with a corresponding enhancement. Then through the conception of this enhancement, we proposed a new approach by going into the distributed domain. Finally, we achieved multiple promising experiments relying on a custom emulator that allowed us to test and measure the performance of our proposals.