Live migration of NoSql databases in the cloud

Details

Supervisors

Riviere, Etienne ; Djeumo Noucti, Carole

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] en sciences informatiques, à finalité spécialisée

Abstract

This thesis explores the challenges and solutions for live migration of NoSQL databases in cloud environments, focusing on the Redis database. Live migration ensures continuous availability, scalability, and efficient resource utilization by dynamically transferring database shards between servers without downtime. The evaluation assesses key performance metrics such as latency, throughput, and consistency under varying workloads and concurrency levels. The results demonstrate that the system performs efficiently under moderate workloads, maintaining low latency and stable throughput. However, challenges arise at higher concurrency levels or with write-intensive workloads, where resource contention and synchronization overhead lead to variability. The limitations of the study include testing with a maximum of 1 million keys, limited concurrency configurations, and the absence of resource limitations such as CPU or bandwidth limitations. The findings highlight the need for improvements, including testing with larger datasets, evaluating diverse workloads, and optimizing mechanisms to track modified keys. By addressing these, the system can become more robust and adaptable, ensuring reliable live migration in dynamic cloud environments.