Serveriprii arvutus värkvõrgu jaoks
Date
2018
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Pilvepõhised teenused on aastate jooksul teinud märkimisväärse arengu. Pilvearvutuse mudelid nagu IaaS, PaaS ja SaaS pakuvad alternatiive traditsioonilistele erataristu-põhistele lähenemistele. Serveriprii arvutus on pilvearvutuse mudel efemeersete, olekuta ja sündmusepõhiste rakenduste jaoks, mis kiirelt skaleeruvad. Vastupidiselt pilvearvutusele, kus ressursse jagub pea lõpmatult, koosneb värkvõrk arvutuslikult piiratud ressurssidega heterogeensetest ja intelligentsetest seadmetest, mis toodavad märkimisväärsetes kogustes andmeid. Värkvõrgu ressursipiiratuse tõttu kasutatakse selliste andmete töötluseks pilveresursse, kuid pilve kasutamine toob kaasa ka mõned piirangud - suurenenud latentsus ning privaatsusprobleemid. Seetõttu tekib vajadus kohalikuks andmetöötluseks värkvõrgu seadmetel. Moodustades tarkvarakonteinerite abil värkvõrgu seadmetel klastri, on võimalik luua serveriprii platvorm kohalikuks andmetöötluseks. Käesolev töö esitleb hübriidset, mitmekihilist arhitektuuri, mis lisaks kohalikule sensorandmete töötlusele arvestab ka probleeme nagu seadmete heterogeensus ja ressursipiiratus. Loodud lahenduses kasutame tarkvarakonteinereid ning mitmekihilist arhitektuuri, et tagada kõrge käideldavus ja rikketaluvus.
Cloud-based services have evolved significantly over the years. Cloud computing models such as IaaS, PaaS and SaaS are serving as an alternative to traditional in-house infrastructure-based approach. Furthermore, serverless computing is a cloud computing model for ephemeral, stateless and event-driven applications that scale up and down instantly. In contrast to the infinite resources of cloud computing, the Internet of Things is the network of resource-constrained, heterogeneous and intelligent devices that generate a significant amount of data. Due to the resource-constrained nature of IoT devices, cloud resources are used to process data generated by IoT devices. However, data processing in the cloud also has few limitations such as latency and privacy concerns. These limitations arise a requirement of local processing of data generated by IoT devices. A serverless platform can be deployed on a cluster of IoT devices using software containers to enable local processing of the sensor data. This work proposes a hybrid multi-layered architecture that not only establishes the possibility of local processing of sensor data but also considers the issues such as heterogeneity, resource constraint nature of IoT devices. We use software containers, and multi-layered architecture to provide the high availability and fault tolerance in our proposed solution.
Cloud-based services have evolved significantly over the years. Cloud computing models such as IaaS, PaaS and SaaS are serving as an alternative to traditional in-house infrastructure-based approach. Furthermore, serverless computing is a cloud computing model for ephemeral, stateless and event-driven applications that scale up and down instantly. In contrast to the infinite resources of cloud computing, the Internet of Things is the network of resource-constrained, heterogeneous and intelligent devices that generate a significant amount of data. Due to the resource-constrained nature of IoT devices, cloud resources are used to process data generated by IoT devices. However, data processing in the cloud also has few limitations such as latency and privacy concerns. These limitations arise a requirement of local processing of data generated by IoT devices. A serverless platform can be deployed on a cluster of IoT devices using software containers to enable local processing of the sensor data. This work proposes a hybrid multi-layered architecture that not only establishes the possibility of local processing of sensor data but also considers the issues such as heterogeneity, resource constraint nature of IoT devices. We use software containers, and multi-layered architecture to provide the high availability and fault tolerance in our proposed solution.