Assessment of Multiple Ingest Strategies for Accumulo Key-Value Store

Abstract

In recent years, the emergence of heterogeneous data, especially of the unstructured type, has been extremely rapid. The data growth happens concurrently in 3 dimensions: volume (size), velocity (growth rate) and variety (many types). This emerging trend has opened a new broad area of research, widely accepted as “Big Data”, which focuses on how to acquire, organize and manage huge amount of data effectively and efficiently. When coping with such Big Data, the traditional approach using RDBMS has been inefficient; because of this problem, a more efficient system named noSQL had to be created. This thesis will give an overview knowledge on the aforementioned noSQL systems and will then delve into a more specific instance of them which is Accumulo key-value store. Furthermore, since Accumulo is not designed with an ingest interface for users, this thesis focuses on investigating various methods for ingesting data, improving the performance and dealing with numerous parameters affecting this process.