Doctoral Thesis Defense - Francisco Miguel Carvalho Barros da Cruz

Date: 
Monday, November 28, 2016, 2:30pm

Title: Towards autonomic workload aware NoSQL databases

Abstract: In order to attain the promises of the Cloud Computing paradigm, systemsneed to transparently adapt to environment changes. NoSQL databases, which are pivotal systems in nowadays cloud infrastructures, exhibit the highly desirable scalability and availability properties. Scalability achieved by these databases is anchored on data independence; there is no clear relationship between data, and atomic inter-node operations are not a concern. Such assumption over data allows a paradigm shift on how to achieve the best performance. Unfortunately, current solutions put the burden on the application’s developer to handle and master the specificities of each system that is hindering a broader adoption. In this dissertation, we tackle the several shortcomings in current implementations of cloud-based NoSQL databases at four diferent levels. First, we present a cloud-enabled framework for the automatic and heterogenous reconfiguration of NoSQL datasses. This framework enables NoSQL databases to become autonomously elastic while providing a new load balancing component that takes into account data access patterns. Secondly, we propose a novel mechanism to partition data that takes into account the system workload. It estimates, in an autonomous way, a splitting point that leads to optimal load balancing in terms of requests. Then, we develop a mechanism to accurately predict the resource usage of NoSQL databases resorting to an offline trained model. It can accurately estimate in real time the database resource usage for any request distribution only by knowing two parameters: i) cache hit ratio; and ii) incoming throughput. This mechanism is sufficiently simple and generic so it can be used with several databases. Finally, we leverage the work on the resource usage prediction to design and implement a novel load balancer mechanism that maximizes the resource usage across the cluster.

Jury: - Rector of the University of Minho

        - Prof. Doctor Paulo Jorge Pires Ferreira

        - Prof. Doctor Gabriel de Sousa Torcato David

        - Prof. Doctor Rui Carlos Mendes Oliveira (Advisor)

        - Prof. Doctor Nuno Manuel dos Santos Antunes

        - Prof. Doctor António Luís Pinto Ferreira de Sousa


Location: Auditorium B2 - CPII - Campus de Gualtar - Universidade do Minho - Braga