Decentralized Self-Adaptation Mechanism for Service Based Applications in Cloud using Spectral Clustering

Cloud computing, with its promise of (almost) unlimited computation, storage and bandwidth, is increasingly becoming the infrastructure of choice for many organizations. As cloud offerings mature, service-based applications need to dynamically recompose themselves, to self-adapt to changing QoS requirements. In this paper, we present a decentralized mechanism for such selfadaptation, using natural language processing. We use a continuous double-auction to allow applications to decide which services to choose, amongst the many on offer. The proposed scheme exploits concepts derived from graph partitioning, and groups together tasks so as to 1) minimize the time overlapping of the tasks assigned to a given resource and 2) maximize the time overlapping among tasks assigned to different resources. The partitioning is performed using a spectral clustering methodology through normalized cuts. Experimental results show that the proposed algorithm outperforms other scheduling algorithms for different values of the granularity and the load of the task requests.