Context-aware composite SaaS using feature model

doi:10.1016/j.future.2019.04.032

2019-10-01

PRE

Jamal Bentahar *

Omar Alam

DOI：10.1016/j.future.2019.04.032

原文链接

原文求助

摘要

En 中文

Software as a Service (SaaS) in cloud computing is delivered in a composite form to effectively address complex levels of user's requirements. Composite SaaS runs in a dynamic distributed cloud environment where the quality of service (QoS) properties of the constituents may get violated at runtime. To face such dynamism and volatility, it is vital to support an online adaptation of composite SaaS. Recent research focused on centralized adaptation approaches based on the closed world assumption that the boundary between SaaS and the cloud environment is known. This is impractical for dynamic composition that requires distributed settings in the open world. To address these challenges, this paper proposes a distributed approach for composite SaaS adaptation applying the master/slave pattern. Slaves locally monitor and adapt the distributed SaaS constituents and send performance information to the master, which adapts the composite service to provide the global expected QoS and monitors the overall performance. To support dynamic adaptation by the master, we propose a solution based on the feature model that captures the variability of the composite SaaS. The activation and deactivation of nodes in the feature model reconfigure the workflow of the composition. Since the reconfiguration task is complex, we apply a meta-heuristic search technique to solve this problem while minimizing the adaptation cost (i.e., resource consumption and violation penalties). Furthermore, we propose an adaption approach for SaaS constituents that substitutes the failed ones promptly to avoid costly global SLA violations. Finally, we present a Kalman-based on-line QoS prediction approach for making decisions regarding the adaptation actions to be taken. Experimental results show that our approach is efficient in distributed and large-scale cloud environments compared to the centralized and off-line approaches. (C) 2019 Elsevier B.V. All rights reserved.

Keyword:

SERVICE

ADAPTATION

SELECTION

ONLINE