Future of Data Management on NoSQL Databases

Gokul Soundarajan (NetApp Advanced Technology Group)




Abstract: NoSQL datastores are a new class of datastores that provide a feature set between those provided by key-value stores and relational databases. There exist different varieties; at one end are simple key-value datastores and other end has columnar databases. The datastores focus on simplicity and speed, often forsaking long-term data management. We instead ask the question: “What capabilities are needed to keep data in NoSQL datastores for the long term?” To answer this question, we look at a typical data lifecycle and study how existing enterprise filesystems handle this question. We survey NoSQL datastores and evaluate them across enterprise data management features. Then, we present FlexDB – an enterprise datastore that uses two key building blocks of snapshots and service-level objectives (topologies) to provide data storage for the long term. 

Bio: Gokul is a researcher at the Advanced Technology Group in NetApp. He received my PhD degree from the University of Toronto in Electrical and Computer Engineering in 2010. His primary interests lie at the intersection of databases and storage. His dissertation looked at methods to run databases on shared storage. His work at NetApp looks at running big-data analytics on shared storage. Gokul has also collaborated with several universities including University of Toronto, University of Wisconsin-Madison, and University of Chicago; he is proud of all the students he has collaborated with.