Probabilistic Quality of Service aware Service Selection

TL;DR

This paper introduces a probabilistic approach to service selection in SaaS environments, modeling QoS profiles with multivariate distributions and assessing their compliance with probabilistic requirements.

Contribution

It presents a novel model for probabilistic QoS profiles, a language for expressing probabilistic requirements, and an automatic method for verification.

Findings

Model effectively captures external influences on QoS values.

Language enables precise specification of probabilistic QoS requirements.

Automatic assessment procedure accurately verifies compliance.

Abstract

In software-as-a-service paradigms software systems are no longer monolithic pieces of code executing within the boundaries of an organisation, on the contrary, they are conceived as a dynamically changing collection of services, collectively executing, in pursuit of a common business goal. An essential aspect of service selection is determining whether the Quality of Service (QoS) profile of a service satisfies the QoS requirements of a client. In realistic execution environments, such QoS values might be influenced by external, non-controllable events, making it impossible for the service provider to guarantee that the values characterised by a QoS profile will be met, naturally leading to the need of a probabilistic interpretation of QoS profile. In this work we propose: 1) a model for describing probabilistic QoS profiles based on multivariate continuous probability distributions, 2) a language for describing probabilistic QoS requirements, and 3) an automatic procedure for assessing whether a probabilistic QoS profile satisfies a probabilistic QoS requirement.