Abstract EN:

Services enable building loosely-coupled and dynamic applications in distributed environments. Service-Level Agreements (SLAs) are used to define service relationships by describing how services should behave. Moreover, SLAs include the Quality of Service (QoS) thal should be delivered along with the service. However, managing service executions on top of distributed infrastructure while meeting agreed QoS is challenging. Firstly, QoS metrics should be mapped to low-level system configurations in order to enable building QoS assurance mechanisms. Secondly, service execution should deal with failures and load variations. Ln addition to these issues, service execution should be driven by pricing aspects since profit increase is an important concern for service providers. Current approaches thal handle service execution support neither the whole SLA life-cycle nor profit augmentation. This thesis proposes an autonomous solution for managing service execution in distributed infrastructures by aiming at increasing the provider profit. Ln particular, this thesis supports the full SLA lifecycle based on: (i) SLA translation under pricing constraints; (ii) mechanisms which ensure fault-tolerant and performance QoS; and (iii) resource acquisition and allocation driven by contract proposais and requests. Ln order to realize this solution, this thesis describes the design and implementation of the Qu4DS (Quality Assurance for Distributed Services) framework. Qu4DS includes a rich set of SLA management functionalities thal provides a higher-level support for service developers. Moreover, Qu4DS is evaluated on top of Grid5OOO and the results show that Qu4DS is able to increase the provider profit while meeting SLAs in different scenarios.