Konteinertehnoloogia ning protsessihaldussüsteemide integratsioonil põhinev adaptiivne protsessijaotus värkvõrgu serval
Files
Date
2017
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Täna levivad pilvepõhised värkvõrgu (asjade interneti) süsteemid tuginevad protsesside halduseks kaugel asuvatel andmekeskustel, mis toob endaga kaasa latentsusprobleeme. Vastusena sellele probleemile on varem välja pakutud servaarvutuse lähenemine, kus arvutused viiakse läbi asjade interneti süsteemi võrgule füüsiliselt lähemal. Mitmete servaarvutuse metoodikate seas on uduarvutus lähenemine, kus rõhk on arvutuste liigutamisel värkvõrgu seadmetele endile. Ehkki uduarvutusel põhinev arhitektuur on paljutõotav, tõstatab see küsimuse – kuidas värkvõrgu protsessihaldussüsteemid (BPMS4IoT-süsteemid) äriprotsesse heterogeensetele värkvõrgu seadmetele jaotama peaksid? Levinud on lähenemine, kus protsesside töövooülesannete käituseks tuginetakse ühisele platvormile. Näiteks, kui haldusserver defineerib teatud töövoo ülesandena Pythoni skripti ja määrab selle seadmele, siis peab seadme töövookäitusmootor toetama vastavat mehhanismi skriptide jooksutamiseks. Selline nõue ei ole paindlik, arvestades värkvõrgu seadmete heterogeensust. Käesolevas magistritöös pakub autor välja raamistiku, mis eraldab töövoo ülesannete käitusmeetodi käitusmootorist kasutades selleks konteinertehnoloogiat. Töö käigus arendati välja raamistiku prototüüp ning viidi läbi katseid mikroarvutitel põhinevail seadmetel. Lisaks võrreldi väljapakutud uduarvutuse raamistiku jõudlust pilvearvutusel põhineva süsteemiga.
Emerging cloud-centric Internet of Things (IoT) system relies on distant data centers to manage the entire processes, which raises the issue of latency. To address the issue, researchers have introduced the Edge computing methodologies that carry out computation closer to the edge network of IoT system. Among the numerous Edge computing approaches, Mist computing paradigm emphasises the mechanism that moves the computation further to the front-end IoT devices. Although the architecture of Mist computing is promising, it raises a new challenge in how the Business Process Management System for IoT (BPMS4IoT) distributes the business process workflow to the heterogeneous IoT devices? In general, executing business process workflows relies on the common platform for executing customized tasks. For example, if the management server defines a Python script task in a workflow, which has been allocated to an IoT device, the workflow engine of the IoT device must have the compatible execution method. Such a requirement is less flexible when one considers the heterogeneity of the IoT devices. Therefore, in this thesis, the author proposes a framework to decouple the workflow task execution method from the workflow engines using the containerization technology. A proof-of-concept prototype has been developed and has been tested on several single-board computers-based IoT devices. Further, a case study has been performed to demonstrate the performance of the proposed framework comparing to the cloud-centric system.
Emerging cloud-centric Internet of Things (IoT) system relies on distant data centers to manage the entire processes, which raises the issue of latency. To address the issue, researchers have introduced the Edge computing methodologies that carry out computation closer to the edge network of IoT system. Among the numerous Edge computing approaches, Mist computing paradigm emphasises the mechanism that moves the computation further to the front-end IoT devices. Although the architecture of Mist computing is promising, it raises a new challenge in how the Business Process Management System for IoT (BPMS4IoT) distributes the business process workflow to the heterogeneous IoT devices? In general, executing business process workflows relies on the common platform for executing customized tasks. For example, if the management server defines a Python script task in a workflow, which has been allocated to an IoT device, the workflow engine of the IoT device must have the compatible execution method. Such a requirement is less flexible when one considers the heterogeneity of the IoT devices. Therefore, in this thesis, the author proposes a framework to decouple the workflow task execution method from the workflow engines using the containerization technology. A proof-of-concept prototype has been developed and has been tested on several single-board computers-based IoT devices. Further, a case study has been performed to demonstrate the performance of the proposed framework comparing to the cloud-centric system.