Publicação

Adaptive Replica Selection in Mobile Edge Networks

Ver documento

Detalhes bibliográficos
Resumo:With the ongoing increase in mobile devices and the application’s growing reliance on the cloud, these infrastructures have become centralized hubs of computational processing and storage. With so much traffic being generated to - and from - these centralized infrastructures, network congestion and delays start to become more evident. Furthermore, having messages travel back and forth to a location that is physically distant from the user severely punishes applications with low latency or high bandwidth demands. Mobile Edge Computing (MEC) is a paradigm that aims to solve these limitations by bringing cloud services closer to mobile clients, effectively reducing end-to-end delays and saving backbone bandwidth. As in a cloud environment, these applications are starting to make use of replication to enhance their quality of service. Because content generated by mobile devices has a localized interest at first, data starts by getting replicated between these devices and only when it starts to get popular is it eventually replicated (cached) in edge servers. The problem arises though, when there is no replica selection mechanism for data retrieval. The resulting herd behavior causes the computational load on the network to be poorly distributed, which combined with the unreliable wireless communication channels cause these systems to under-perform. In thesis we propose Wasabi, an adaptive replica selection algorithm for MEC environments with the aim of decreasing latency and boosting both throughput and energy efficiency in MEC systems. Furthermore, we develop a whole replica selection framework to support Wasabi and its integration with Thyme GardenBed [14]. From our experimental results, we conclude that Wasabi performs better in dynamic environments than any of the presented baselines, including the cloud algorithm C3 [17] and its MEC variant, which make use of a similar set of metrics.
Autores principais:Dias, João Pedro Monteiro Morgado
Assunto:Mobile Edge Computing Replica Selection Mobile-to-mobile Mobile-to-edge
Ano:2021
País:Portugal
Tipo de documento:dissertação de mestrado
Tipo de acesso:acesso aberto
Instituição associada:Universidade Nova de Lisboa
Idioma:inglês
Origem:Repositório Institucional da UNL
Descrição
Resumo:With the ongoing increase in mobile devices and the application’s growing reliance on the cloud, these infrastructures have become centralized hubs of computational processing and storage. With so much traffic being generated to - and from - these centralized infrastructures, network congestion and delays start to become more evident. Furthermore, having messages travel back and forth to a location that is physically distant from the user severely punishes applications with low latency or high bandwidth demands. Mobile Edge Computing (MEC) is a paradigm that aims to solve these limitations by bringing cloud services closer to mobile clients, effectively reducing end-to-end delays and saving backbone bandwidth. As in a cloud environment, these applications are starting to make use of replication to enhance their quality of service. Because content generated by mobile devices has a localized interest at first, data starts by getting replicated between these devices and only when it starts to get popular is it eventually replicated (cached) in edge servers. The problem arises though, when there is no replica selection mechanism for data retrieval. The resulting herd behavior causes the computational load on the network to be poorly distributed, which combined with the unreliable wireless communication channels cause these systems to under-perform. In thesis we propose Wasabi, an adaptive replica selection algorithm for MEC environments with the aim of decreasing latency and boosting both throughput and energy efficiency in MEC systems. Furthermore, we develop a whole replica selection framework to support Wasabi and its integration with Thyme GardenBed [14]. From our experimental results, we conclude that Wasabi performs better in dynamic environments than any of the presented baselines, including the cloud algorithm C3 [17] and its MEC variant, which make use of a similar set of metrics.