SLA management of non-computational services

Abstract

The rise of computational services in the last 15 years brought the proposal of a number of techniques to automate and support their enactment. One key element in services is the Service Level Agreement (SLA), where the requirements of service customer are matched with the performance levels from the service provider to define service level guarantees and related responsibilities. The proposals from computational domains are oriented to automate the different stages in the SLA Lifecycle, such as the negotiation of terms which will form the SLA, the deployment of services based on the SLA artifact or the management of computational resources to accomplish SLA goals on runtime. However, traditional non-computational services, that is, services which are not performed by computational resources, such as logistics or software development services, are still supported by ad-hoc mechanisms. Therefore, the existing solutions for the management of their SLAs cannot be reused for other services. This management is usually manually performed (e.g.: reviewing of the goals of an SLA in transport service), so their evaluation is error-prone and delayed regarding the service execution (e.g.: when the SLA is finished), so preemptive actions to avoid SLA violations cannot be taken or/and are expensive to perform. Furthermore, these SLAs are sometimes described on a long term basis (frame agreements), and related SLAs can appear for a shorter term (specific agreements) and the analysis of the validity among them is complex to perform on runtime. In this dissertation, we aim at partially automate the management of SLAs in noncomputational services. On the one hand, we suggest that existing models for computational services can be extended to non computational services and enable the description of the service operative and their guarantees. And, on the other hand, we provide a design for operations to partially support the SLA Lifecycle, based on the previous models. Specifically, these operations are mainly focused on the deployment and fulfillment stages of the SLA. Therefore, the contributions of this dissertation are three. First, (A) providing a model to describe Service Level Agreements of non computational services, as an extension of iAgree, an existing model for SLAs of computational services. Second side, (B) supporting the SLA Lifecycle with the design of the aforementioned operations (service configuration based on SLA and monitoring of SLA) and implementing a reference architecture for such operations. And, lastly, (C) providing a model for frame and specific agreements which relates their terms and formalises the analysis operations among them. Other related operations of the service lifecycle as the management of resources to improve service performance or the use of novel techniques (such as machine learning) to predict the SLA accomplishment are out of the scope of this thesis but planned as future line of extension. The current dissertation has been based on real SLAs from different domains, such as Transport & Logistics, public Cloud providers or IT Maintenance outsourcing, which have been used to validate the proposal. And, furthermore, the contributions have been applied in the context of real IT Maintenance outsourcing projects.