Formal Quality of Service assurances, ranking and verification of cloud deployment options with a probabilistic model checking method

Abstract

Context: Existing software workbenches allow for the deployment of cloud applications across a variety of Infrastructure-as-a-Service (IaaS) providers. The expected workload, Quality of Service (QoS) and Non-Functional Requirements (NFRs) must be considered before an appropriate infrastructure is selected. However, this decision-making process is complex and timeconsuming. Moreover, the software engineer needs assurances that the selected infrastructure will lead to an adequate QoS of the application. Objective: The goal is to develop a new method for selection of an optimal cloud deployment option, that is, an infrastructure and configuration for deployment and to verify that all hard and as many soft QoS requirements as possible will be met at runtime. Method: A new Formal QoS Assurances Method (FoQoSAM), which relies on stochastic Markov models is introduced to facilitate an automated decision-making process. For a given workload, it uses QoS monitoring data and a user-related metric in order to automatically generate a probabilistic model. The probabilistic model takes the form of a finite automaton. It is further used to produce a rank list of cloud deployment options. As a result, any of the cloud deployment options can be verified by applying a probabilistic model checking approach. Results: Testing was performed by ranking deployment options for two

Trust management in a blockchain based fog computing platform with trustless smart oracles

“…TensorFlow). This layer uses a Markov probabilistic decision-making method for automated decision-making [58]. In order to rank the infrastructures, the Markov method requires QoS monitoring data and QoS threshold values from a specially designed Smart Oracle.…”

Smart Contracts for Service-Level Agreements in Edge-to-Cloud Computing

“…This layer is composed of components that are products of our earlier research work. In particular, the Decision-Making Layer is composed of three systems: decision-making mechanism [15], monitoring system [31] and an orchestration system [22]. The implemented decision-making mechanism is based on the Markov Decision Process (MDP) that generates a probabilistic finite automaton that is built for each microservice.…”

“…Transition probabilities, which are estimated from prior usage experience of the deployment options and state rewards, which are estimated from the monitoring metrics are essential when calculating the utility of each state. A detailed description of the algorithm including the calculation of rewards and transition probabilities is available elsewhere [15].…”

“…In addition to the above, our novel architecture utilises resource provisioning and application placement methods, which are based on Markov Decision Process (MDP) to model both internal and external (contextual information) metrics that affect the resulting QoS of the individual AI applications [15]. Since the runtime conditions can dynamically change, an automaton -a model derived by using an MDP -can be used in order to obtain QoS assurances, ranking and verification of available deployment options.…”

Model Checking Based Approach for Compliance Checking

“…They propose an approach based on Petri Nets and illustrate its functioning through a simple example related to an access control system. A Formal Quality of Service Assurances Method which relies on stochastic Markov models is proposed in [15] with the aim to facilitate the decision-making process. They consider probabilistic model checking with a set of user-related metric to automatically generate a probabilistic model.…”