Andreea Buga

Towards a Case-Based Reasoning Approach to Dynamic Adaptation for Large-Scale Distributed Systems

The ever growing demands from the software area have led to the development of large-scale distributed systems which bring together a wide pool of services and resources. Their composition and deployment come in different solutions tailored to users requests based on business models, functionality, quality of service, cost, and value. Bridging different parts into one software solution is brittle due to issues like heterogeneity, complexity, lack of transparency, network and communication failures, and misbehavior. The current paper proposes a decision-based solution for the dynamic adaptation part of a middleware which addresses the aforementioned problems for large-scale distributed systems. The envisioned architecture is built on case-based reasoning principles and stands at the base of the adaptation processes that are imperative for ensuring the delivery of high-quality software. The solution is further extended through ground models with a focus on reliability, availability of components, and failure tolerance in terms of abstract state machines. The novelty of the approach resides in making use of formal modeling for one of the emerging problems and introducing an adequate prototype, on top of which one can apply reasoning and verification methods.

Towards Modeling Adaptation Services for Large-Scale Distributed Systems with Abstract State Machines

The evolution of Large-Scale Distributed Systems favored the development of solutions for smart cities. Such systems face a high-level of uncertainty as they consist of a large number of sensors, processing centers, and services deployed along a wide geographical area. Bringing together different resources poses increased complexity as well as communication efforts, and introduces a large set of possible failures and challenges of continuously growing computational and storage expectations. In such a frame, the role of the adaptation components is vital for ensuring availability, reliability, and robustness. This paper introduces a formal approach for modeling and verifying the properties and behavior of the adaptation framework addressing the case of a system failure. We formalize the behavior and the collaboration mechanisms between agents of the system with the aid of Abstract State Machines and employ the ASMETA toolset for simulating and analyzing properties of the model.

Conceptual Modelling of Autonomous Multi-cloud Interaction with Reflective Semantics

Distributed systems that exploit software services from multiple clouds provide opportunities for software systems that address problems associated with systems of systems. In this paper we present an approach for the conceptual modelling of such systems, which is grounded in a distributed middleware that coordinates the client access to multiple clouds through a concept of mediator. Furthermore, each component of the middleware constitutes an abstract machine that is realised by three layers: a layer for normal operation, a layer for monitoring and detection of critical situations, and an adaptation layer, which in case of an identified anomaly changes the normal behaviour. The semantics of this autonomous system can be captured by linguistic reflection, for which reflective Abstract State Machines will be exploited.

A Formal Approach for Failure Detection in Large-Scale Distributed Systems Using Abstract State Machines

Large-scale distributed systems have been widely adopted in various domains due to their ability to compose services and resources tailored to user requirements. Such systems are characterized by high complexity and heterogeneity. Maintaining a high-level availability and a normal execution of the components implies precise monitoring and robust adaptation. Monitors capture relevant metrics and transform them to meaningful knowledge, which is further used in justifying adaptation actions. The current paper proposes an Abstract State Machine model for defining monitoring processes addressing failures and unavailability of the system nodes. The specification is simulated and validated with the aid of the ASMETA toolset. The solution is complemented with a small ontology reflecting the structure of the system. We emphasize the role of formal models in achieving the proposed requirements.

AsthMate -- Supporting Patient Empowerment through Location-Based Smartphone Applications

The ever changing challenges and pressures to the healthcare domain have introduced the urgency of finding a replacement for traditional systems. Breakthroughs registered by information systems, advances in data storage and processing solutions sustained by the ubiquity of gadgets and an efficient infrastructure for network and services have sustained a shift of medical systems towards digital healthcare. Asth Mate application is an e-health tool for asthma patients, acting as an enabler for patient empowerment. The contributions brought by the application are both to the individual and to the community exposing a web application that allows citizens to check the state of the air for the area they live in. The ongoing implementation can benefit of the advantages of cloud computing solutions in order to ensure a better deployment and data accessibility. However, data privacy is a key aspect for such systems. In consideration of this reason, a proper trade-off between the functionality, data openness and security should be reached.

An Event-B-based approach to hybrid systems engineering and its application to a hemodialysis machine case study

Abstract Systems engineering concerns the complete process for the development of complex systems comprising hardware, software, facilities and personnel. Such systems are hybrid, as some components are characterized by continuous behavior, whereas the behavior of others is discrete. In this paper we present a concise conceptual model for hybrid systems engineering with semantics grounded in a hybrid extension of Event-B. We show that structural modeling can be based on well-known concepts of the entity-relationship model requiring only some extensions to data types and constraints, while behavioral modeling requires a careful separation of synchronous and asynchronous interaction and high-level means for the integration of continuous functions. On these grounds we address the separation of concerns for continuous and hybrid components. The article uses a sophisticated industrial example of a hemodialysis machine to illustrate the modeling method.

Conceptual Modelling of Hybrid Systems

Complex systems comprising hardware, software, facilities and personnel are gaining more and more importance. Such systems are hybrid, as some components are characterised by continuous behaviour, whereas the behaviour of others is discrete. In this paper we present a concise conceptual model that is capable to capture structure and behaviour of such systems. We show that structural modelling can be based on well-known concepts of the entity-relationship model requiring only some extensions to constraints. We further show that behavioural modelling requires only a careful separation of synchronous and asynchronous interaction and high-level means for the integration of continuous functions. We show that all these concepts can be captured by defining a semantics in hybrid Event-B. The paper illustrates the modelling method by a sophisticated industrial example of a hemodialysis machine.

Formalizing Monitoring Processes for Large-Scale Distributed Systems Using Abstract State Machines

Large-Scale Distributed Systems are characterized by high complexity and heterogeneity, which might lead to unexpected failures. The role of a robust monitoring framework is to gather low-level data and assess the status of the components of the system. The framework collaborates with adapters for ensuring steady recovery plans and improving the availability of services. Monitors, as part of the system, are also affected by unavailability or random failures. In order to increase the reliability of the solution we verify the trustworthiness of the monitors and emphasize the need of redundancy. This paper introduces a formal approach for modeling and verifying a monitoring solution for Large-Scale Distributed Systems. We formalize the behavior of the monitors with the aid of Abstract State Machines and employ the ASMETA toolset for simulating and analyzing properties of the model. The tool also supports the verification process by translating a simplified version of the model to an NuSMV specification, on top of which model checking can be applied. Properties of the model are expressed with the aid of computation tree logic.

Environment crowd-sensing for asthma management

Chronic diseases are the main cause of mortality in developed countries. Among them, asthma has an increasing prevalence in urban areas. We propose to use the crowd-sensing paradigm for monitoring the environment, especially the air quality parameters relevant for asthma patients. Asthma is a good candidate for participatory medicine approaches as current medical guidelines promote self-management of the disease by informed and ICT enabled patients. The paper continues our work for m-health applications for asthma patients and focuses on decentralized air quality monitoring by the patients themselves using affordable and available hardware platforms based on Arduino.

Towards Care Systems Using Model-Driven Adaptation and Monitoring of Autonomous Multi-clouds

In cloud computing, the ability to run and manage multi-cloud systems allows exploiting the peculiarities of each cloud solution and hence optimising the performance, availability, and cost of the applications. In this paper, we investigate the use case of a robotic care system as an application of autonomous multi-clouds. We present requirements and properties of an Abstract State Machines-based conceptual model that coordinates the multi-cloud interaction through the specification of a middleware exploiting adaptive interfaces to multiple clouds and supporting various service formats. While the multi-cloud system is running in normal mode, data about the execution will be gathered and evaluated by the monitoring component, and in case any critical situation is discovered the adaptation component is alerted. We show that for the care system this can be fruitfully exploited for failure alerts, failure anticipation and prevention, and safety hazards detection.