A middleware for managing dynamic software adaptation

Abstract

The design and development of adaptive systems brings new challenges since the dynamism of such systems is a multifaceted concern that range from mechanisms to enable the adaptation on the software level to the (self-) management of the entire system using adaptation plans or system administrator, for instance. Networked and mobile embedded systems are examples of systems where dynamic adaptation become even more necessary as the applications must be capable of discovering the computing resources in their near environment. While most of the current research is concerned with low-level adaptation techniques (i.e., how to dynamically deploy new components or change parameters), we are focused in providing management of distributed dynamic adaptation and facilitating the development of adaptation plans. In this paper, we present a middleware tailored for mobile embedded systems that supports distributed dynamic software adaptation, in transactional and non-transactional fashion, among mobile devices. We also present results of initial evaluation.

Dynamic and coordinated software reconfiguration in distributed data stream systems

“…Whenever a component's parameter is updated, their work uses a state transfer mechanism to perform an update from the old component state to the new one. However, the proposal does not allow the modification of a component by its new version (i.e., it does not permit compositional adaptation [10,46]). Finally, the systems considered by the authors are always deployed on a single node (i.e., a car), unlike ours where the system is distributed and has multiple instances of a component type.…”

An Adaptive Middleware for Opportunistic Mobile Sensing

“…Our adaptive approach provides a simple development model, which is based on components, that hide most of the complexity of dynamism management, transparently loading, unloading and updating the M-Hub's sensor modules. Moreover, our approach supports distributed and transactional software adaptations among M-Hubs by using adaptation plans (i.e., sequences of commands that are executed at the M-Hubs), which are described in [7] [8]. However, considering the M-Hub's functionalities presented in this paper, we focus only on some local adaptations performed at a single M-Hub.…”

“…Most of our implementation uses software engineering designs (e.g., interfaces and abstract classes) and programming techniques (e.g., Java generic types) in order to reduce the use of computational reflection, since it introduces a considerable overhead at runtime [11], [12]. More details about the adaptive layer are found in [7] [8]. From a global perspective, we assume that extensions and modifications to the set of sensors used in a crowd-sensing application using the M-Hub are to be initiated by the provider or owner of the new/updated sensor devices, as follows: he/she will submit the sensor device specs to the developer/operator of the crowd-sensing application, who will then develop the transcoding module and its wrapper, and produce corresponding metadata about the new sensor (e.g.…”

Middleware for Customizable Multi-staged Dynamic Upgrades of Multi-tenant SaaS Applications

“…While their support is based on types of applications that differs from multi-tenant SaaS application, their work is complementary to ours as it focuses on an algorithm for automated enactment (corresponds to our activation mechanism), accounting for state-transfer, referential integrity and timeliness of dependent upgrades. In the context of the Internet of Things, a middleware has been proposed for dynamic adaptation in the sense that collaboration with unanticipated services is supported seamlessly [50]. While we focus on service continuity improvements, their focus is on service discovery and ad-hoc interaction.…”