Big data systems are stable enough to store and process large volumes of quickly changing data. However, these systems are composed of massive hardware resources, which can easily cause their subcomponents to fail. Fault tolerance is a key attribute of such systems as they maintain availability, reliability and constant performance during failures. Implementing efficient fault-tolerant solutions in big data presents a challenge because fault tolerance has to satisfy some constraints related to system performance and resource consumption. To protect online computer systems from malicious attacks or malfunctions, log anomaly detection is crucial. This paper provides a new approach to identify anomalous log sequences in the HDFS (Hadoop Distributed File System) log dataset using three algorithms: Logbert, DeepLog and LOF. Then, it assess performance of all algorithms in terms of accuracy, recall, and F1-score.

One key perspective when communicating with deaf people is sign language recognition. Many experts agree upon the fact that using a system communication is a key for bridging the gap between deaf and non deaf people since ordinary people can not exchange using the sign language. As a result, researchers, deaf people, parents and deaf-mute community are striving to have a bidirectional communication system based on translation with lower costs. The importance of the research is related to its goal of helping these categories of unvoiced people communicate with others and enhance their contributions to growth and capacity building and vice versa.This paper gives an overview of the most used techniques and technologies (gloves, android application, image processing, …) in order to translate sign language to written or spoken language. Furthermore, this paper provides a critical and comparative analysis of the studied approaches and stands out major challenges to overcome their limits. Finally, we propose in this paper a dual way communication system ensuring arabic sign language translation into spoken language based on image processing and deep learning.

AADL (Architecture Analysis & Design Language) is a modelling language enabling the design of embedded realtime systems. This language has proven its power in several domains like aeronautics and space. The AADL core language describe both software and hardware components and it can be extensible by properties to provide information about these components or annexes to enable a designer to extend and customize the AADL core specification with other concepts specified in a language other than AADL. In particular, the AADL Error Model Annex (EMA) was proposed to model such requirements separately from the core model since dependability requirements is of major importance in real-time embedded systems. This annex can not only model different kinds of errors but also error propagation, error detection and error recovery. In this paper, we ensure a comparative study of existing tools and compilers of the AADL language and its error model annex. We present, then, our work aiming at integrating a new compiler of the EMA annex into the Ocarina tool suite to support dependability requirements.

Aspect-Oriented Programming (AOP) supports the separation of concerns cutting across the implementation of a software application in order to avoid scattering and tangling. We focus on the integration of this paradigm into the development of real-time systems. AOP is based on the weaving process to automatically integrate separately-created segments of code (specified by aspects), into the application by replacing them in the final implementation code. In the context of real-time systems, this can harm determinism by including constructs that may violate real-time constraints beyond the control of the developer such as dynamic allocation. We both studied and tested the AspectAda language, an existing aspect mechanism, in order to adapt its implementation to respect real-time constraints. We studied this language's grammar, runtime and especially weaving operation. In this paper, we briefly discuss both the limitations of the existing language and the proposed solutions. Then, we present a new architecture of the compiler/weaver to overcome the problems of the existing one and we detail the implementation of its different parts based on tree manipulation. Finally, we illustrate our contribution using a case study.

Fault tolerance, one of the means of dependability, is defined as the capability of a system to continue providing offered services even in the presence of errors. Modelling fault tolerant systems is becoming a serious challenge due to emerging requirements suh as dynamism, distribution and system evolution. In this context, AADL is extended by the Error Model Annex to design architectural error model even for critical systems for analysis and assessment purposes. In this paper, we define a model driven approach to generate fault tolerant application code using aspect oriented programming. First, we integrate fault tolerance elements since the modelling step of a system using AADL and its annex Error Model Annex. Second, we define a set of transformation rules from the EMA annex to the AO4AADL annex, an extension of the AADL language by architectural aspects. The model transformation is generic and not specific to a language. AO4AADL is transformed into aspect code using the same target language as the functional code. Therefore, we ensure separation of concerns at the model level as well as the application level. To validate our approach, we apply it to an infant incubator system.