Leveraging metadata in NoSQL storage systems

We present Wasef, a metadata system for NoSQL data stores. Wasef allows us to support important features such as data provenance, flexible data operations, and new administration tools. Our system has a general design that can be made compatible with most NoSQL data stores without imposing heavy performance overhead. We implement a customized version of the Cassandra data store augmented with Wasef, and we compare it experimentally with the standard version. We also provide three use-case scenarios that exploit the metadata collected by our system. First, we support a flexible mechanism for dropping columns that addresses an open major issue in Cassandra 1.2. Second, we make provenance information for Cassandra queries available to the data store clients. Third, we provide a tool for verifying node decommissioning from Cassandra cluster. Our evaluation shows that Wasef provides linear scalability, low throughput overhead of 9%, and read latency overhead of less than 3% compared to the non-metadata default Cassandra. In addition, our system provides steady read and update latencies regardless of the metadata size.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2014

Author dc:creator

• Alkhaldi, Ala'

Contributors dc:contributor

• Gupta, Indranil

• Metadata
• Not only Standard Query Language (NoSQL)
• Cassandra

Statement dc:rights

• Copyright 2014 Ala' Alkhaldi

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/49626

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/49626