Volume 56, Issue 1

In an increasing number of use cases, databases face the challenge of managing heterogeneous data. Heterogeneous data is characterized by a quickly evolving variety of entities without a common set of attributes. These entities do not show enough regularity to be captured in a traditional database schema. A common solution is to centralize the diverse entities in a universal table. Usually, this leads to a very sparse table. Although today's techniques allow efficient storage of sparse universal tables, query efficiency is still a problem. Queries that address only a subset of attributes have to read the whole universal table including many irrelevant entities. A solution is to use a partitioning of the table, which allows pruning partitions of irrelevant entities before they are touched. Creating and maintaining such a partitioning manually is very laborious or even infeasible, due to the enormous complexity. Thus an autonomous solution is desirable. In this article, we define the Online Partitioning Problem for heterogeneous data. We sketch how an optimal solution for this problem can be determined based on hypergraph partitioning. Although it leads to the optimal partitioning, the hypergraph approach is inappropriate for an implementation in a database system. We present Cinderella, an autonomous online algorithm for horizontal partitioning of heterogeneous entities in universal tables. Cinderella is designed to keep its overhead low by operating online; it incrementally assigns entities to partition while they are touched anyway during modifications. This enables a reasonable physical database design at runtime instead of static modeling.

Database design is a well-studied and well-understood problem in developing database applications. However, modern application requirements in terms of query response time, query throughput, database sizes, database maintenance and database interactivity capabilities on the one hand, and the manifold features of modern DBMS for physical schema design such as numerous different index structures, materialized views, partitioning, and clustering schemes on the other hand make high demands on the database design process. With this growing number of impact factors on physical database design automatic solutions have been developed and are provided in most systems, while research in this area still continues. In this paper, we give a survey on the state of the art in autonomic physical design. Starting with the classic index tuning problem and possible solutions, we describe further design problems such as choosing materializations of aggregations for OLAP and multidimensional clustering schemes. Finally, we point out to most recent challenges resulting from a shift in requirements and give a direction on how to tackle these.

In today's database server environments, multiple types of workloads can be present in a system simultaneously. Workload types may include on-line transaction processing and business intelligence. Workloads may also have different levels of business importance and distinct performance objectives, which are typically derived from service level agreements. An autonomic workload management system for database management systems (DBMSs) dynamically monitors and controls the flow of the workloads to help DBMSs achieve the desired performance objectives. In this paper, we present a framework and a prototype implementation for autonomic workload management in DBMSs. The framework and the prototype provide the ability to achieve performance objectives of workloads with diverse characteristics, different levels of business importance and varying resource demands while protecting DBMSs against performance failure. The prototype system is implemented on top of IBM ® DB2 ® Workload Manager. Initial experiments using the prototype system are presented to demonstrate the effectiveness of the framework.

Ubiquitous computing applications provide pervasive support to users in a self-adaptive and non-obtrusive way. Applications reason about the user situation and adapt dynamically, often without explicit user interaction. Such applications exploit technical features such as context awareness, context reasoning, adaptation models, dynamic resource discovery/binding, and self-configuration. The engineering of ubiquitous computing applications is a challenge because the user acceptance depends not only on functional features but at least as much on non-functional and user-related features that we address under the term “social awareness”. In this paper we present an interdisciplinary development approach for self-adaptive applications that takes into account social awareness requirements in a systematic and integrated manner. Our focus is on usability, trust, and legality. We present the ingredients of our new methodology and its evaluation based on application prototypes. Our approach is compatible with existing software engineering process models and practices.