Migration in Edge Computing

Arshin Rezazadeh, Western University

Abstract

Mobile IoT applications often require low response time and high bandwidth. These applications include virtual reality, augmented reality, and online gaming. Currently, most data processing is done in the cloud. However, for latency-sensitive applications, the latency may need to be reduced. Edge and fog computing can be used to place application services close to mobile devices to reduce latency. However, as mobile devices move, latency increases, which can be decreased by moving the service to a closer edge/fog server. This can be addressed by migrating services so that the mobile device can receive services from the new server. These services can be run on a single or multiple virtual machines or containers.

Application services must be migrated together. Delays in multi-service migration can occur because some application services migrate faster than others, requiring them to wait for the rest of the services to migrate before continuing their tasks. However, in the last decade, most migration research has focused on cloud computing rather than edge and fog computing. Furthermore, migration methods and models are primarily intended for cloud applications.

This thesis focuses on migration techniques for managing IoT applications in edge and fog computing environments while considering the characteristics of IoT applications, the networking characteristics of the edge and fog servers, and the migration parameters of running IoT applications. This thesis contributes to the current state of the art by presenting the following contributions in fog and edge computing environments:

1. A comprehensive literature review and limitations on the migration of IoT applications from different perspectives, namely downtime and migration time reduction strategies, optimization techniques, and identifying critical parameters of migration.
2. A new migration method for latency-sensitive IoT applications to reduce downtime and migration time, including performance evaluations and comparisons to well-known migration methods.
3. New comprehensive models with non-average parameter values for higher precision and accuracy, including comparative analysis of the parameters that impact the performance of the investigated migration methods.
4. A new bandwidth allocation strategy for multi-service migrations of IoT applications to reduce downtime and migration time, including performance evaluations.