UdS Infosys

OctopusDB

This page describes my contributions to OctopusDB. See the detailed webpage at Saarland University here.

The last decade has seen a proliferation of software systems for storing and managing large collections of data. These different systems are targeted at different types of data, or different data access patterns; examples include: transactional applications, such as banking (OLTP), analytical and reporting applications, such as finding the most popular items that have sold in a retail store (OLAP), as well as systems specialized for streaming (real-time), social networks and other graphs, and data archives. This is a sharp change from just a few decades ago where all applications were supported by monolithic data management systems (e.g., Oracle, or DB2). This proliferation of systems has occurred because it delivers significantly better performance. However, end users now need to pick the right data management system for their query workloads.

Furthermore, modern enterprises typically see a variety of query workloads. For instance, a banking enterprise uses a transactional system for customer banking transactions, an analytical system for business intelligence, a streaming system for stock trading, and an archival system to meet regulatory requirements for data retention. As a consequence, today's companies have to manage and integrate several types of data management systems, which is tedious, expensive, and counter-productive for their business.

Our vision is a highly flexible data managing system, which automatically sets the initial configuration and adapts to changing workload later on, with improved performance, lowered cost and better maintainability. This vision for a new system, coined OctopusDB, initially stores a log or journal of data operations. Thereafter, depending on the workload, OctopusDB creates arbitrary physical representations, called Storage Views, of that journal. As a result of this flexible data storage layer, OctopusDB can mimic a variety of systems and efficiently support dynamic query workloads.


Publications
Talks and Posters
Patent
  • Jens Dittrich, Alekh Jindal
    A method for storing and accessing data in a database system.
    US Patent US20130226959 A1