With Great Freedom for Inconsistent Data Comes Great Scalability Responsibility

Shared online databases, such as Google Fusion Tables or Quickbase, allow community members to collaboratively maintain and browse data. While users may believe in conflicting facts (due to conflicting sources, measurements or opinions), current online databases do not offer support for the management of data conflicts. Thus online databases could clearly benefit from technology for uncertain/incomplete databases. However, prior works on uncertain databases are of limited help when designing a conflict-aware online database, for two reasons: First, their performance degrades rapidly as the number of conflicting facts escalates, which can be the case in large user communities. Second, they were built as storage models, resulting in data models that are either non-simple or non-compact and thus may require additional, often non-trivial processing before they appear in the Frontend of an online database. To overcome these problems, we describe Ricolla; a scalable online database with built-in support for data conflict management. Ricolla allows users to model conflicting data, inspect them in a compact form and resolve inconsistencies in an “as-you-go" personalized fashion, even in the presence of a large number of conflicts. It achieves this by coordinating a novel conflict-aware data model (shown to be both compact and simple) with respective efficient query answering algorithms, that allow the system to scale to a large number of data conflicts (as evidenced by a performance comparison against Trio and MayBMS; two recent research prototypes for uncertain data). In parallel, the data model’s simplicity and compactness make it suitable for direct use by the Frontend.