Quality of service-aware co-engineering of cloud applicationsŠtefanič, Polona (Avtor) Stankovski, Vlado (Mentor) Fog and Cloud ComputingInternet of ThingsMicroservicesSubgraph IsomorphismAdaptationApplications that make use of Internet of Things (IoT) capture an enormous amount of raw data from fixed and dynamically located sensors and devices. Currently, IoT devices send data directly to the cloud systems for processing to benefit from high availability, scalability, unlimited storage and pooled computing resources on the pay-per-use models. Due to ever-increasing and unpredictable data generation rates and varying demands of Iot-based applications on compute resources, sending the data towards the clouds can result in high latency, transmission costs and privacy issues which cannot efficiently support the overall Quality of Service (QoS) of latency-critical microservice-oriented applications. Recently, the emergence of Fog and Edge computing has enabled to move the services, data and processing power towards the network edge. However, there is lack of QoS-aware approaches that would fully address the QoS-aware reconfigurability of on-demand compute resources and service placement along Cloud-To-Things Computing Continuum. For this purpose, we propose a new P-Match method for QoS-aware distributed deployment and continuous adaptation of interactive and latency-critical applications composed of microservices. The P-Match method as an input data considers metrics obtained from virtual infrastructure monitoring and returns a set of QoS-aware infrastructure deployment options for the deployment of microservice-oriented applications. The P-Match method utilises a subgraph isomorphism matching that is based on requirements and resource matchmaking of an applications' components constraints towards the multi-level QoS metrics related to the Cloud-To-Things Continuum. The overall comparison of the P-Match method with MDP method also suitable for continuous adaptation reveals that for the particular set of microservices' constraints both methods proposed continuous deployment that satisfied the given QoS constraints. Though both methods gave QoS-aware solutions and satisfied the requirements, P-Match returned four times less powerful deployment options regarding infrastructure-level metric and consequently made more sustainable decision yet still enough to satisfy the particular QoS constraints. Since P-Match reached the same or better solutions than MDP method it obviously proves its correctness. Due to its lightweight nature, the P-Match method is suitable for the integration into service modelling frameworks and can be utilised particularly in the provisioning phase to support service placement, tasks and jobs scheduling and execution to specific (virtual) instances to satisfy application's QoS and contributes to the sustainable usage of compute resources, adjusts latency of microservices based on their requirements and data processing at instances where it is the most meaningful.20212021-08-16 15:10:00Doktorsko delo/naloga128915VisID: 23162COBISS_ID: 74679043sl