Improving service quality in cloud computing : from definition to deployment

Service quality is crucial in all stages of the Cloud service life cycle, from service acquisition, where Cloud consumers and providers negotiate for a mutual agreement, to service execution, where service management is driven by the agreed requirements. Much work has been devoted to specification and enforcement of service quality terms in the Cluster, Grid and Cloud domains. However, the dynamism present in Cloud services is ignored. We propose a theoretical and practical framework which addresses the first phases of the service life cycle: (i) the definition of service provision; (ii) the negotiation of offers/requests expressed and (iii) the service deployment, mainly focused on latency-sensitive applications. We introduce SLAC, a specification language for the definition of service requirements, the so-called service level agreements (SLAs), which allows us to define conditions and actions that can automatically modify those terms at runtime. Experimental results show that the use of SLAC can drastically reduce the service violations and penalties to the advantages of providers and consumers. Then, we define a novel matchmaking and negotiation framework, which evaluates the compatibility of SLAC requests/offers, and provides the modifications necessary to reach an agreement. Experiments demonstrate the effectiveness of our proposal. We also introduce a new scheduling algorithm for latency-sensitive services, in a Cloud/Edge computing scenario, which takes into account not only the service requirements but also network latency, bandwidth and computing capabilities. Again, experimental results confirm the advantages of this new approach over existing solutions.