Open Collections

Scalable Database Management System (DBMS) architecture with Innesto

University of British Columbia

Database Management systems (DBMS) have been in the core of Information Systems for decades and their importance is getting higher and higher with current high growth in user demand and rising necessity to handle big data. With recent emergence of new style of deployments in the cloud, decades old architectures in DBMS have been greatly challenged due to their inability to scale beyond single computing node and to handle big data. This new requirement has spawned new directions along scaling data storage architectures. Most of the work surfaced lacks the applicability across many domains as they were targeting only a specific domain. We present a novel scalable architecture which is implemented using a distributed spatial partitioning tree (SPT). This new architecture replaces only the storage layer of a conventional DBMS thus leaving its applicability across domains intact and provides strict consistency and isolation. Indexing and locking are two important components of a Relational Database Management System (DBMS) which pose as potential bottleneck when scaling. Our new approach based on SPT provides a novel scalable alternative for these components. Our evaluations using the TPC-C workload show they are capable of scaling beyond single computing node and support more concurrent users compared to a single node conventional system. We believe our contributions to be an important first step towards the goal of a scalable, cloud aware and full-featured DBMS as a service.

Text

2012-10-10

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/43363

Computer Science

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/