Data warehouse queries pose challenging performance problems that often necessitate the use of parallel database systems (PDBS). Although dynamic load balancing is of key importance in PDBS, to our knowledge it has not yet been investigated thoroughly for parallel data warehouses. In this study, we propose a scheduling strategy that simultaneously considers both processors and disks while utilizing the load balancing potential of a Shared Disk architecture. We compare the performance of this new method to several other approaches in a comprehensive simulation study, incorporating skew aspects and typical data warehouse features such as star schemas.

So far only few performance studies on parallel object-relational database systems are available. In particular, the relative performance of relational vs. reference-based join processing in a parallel environment has not been investigated sufficiently. We present a performance study based on the BUCKY benchmark to compare parallel join processing using reference attributes with relational hash- and merge-join algorithms. In addition, we propose a data allocation scheme especially suited for object hierarchies and set-valued attributes.

Skew effects are a serious problem in parallel database systems, but the relationship between different skew types and load balancing methods is still not fully understood. We develop and compare two classifications of skew effects and load balancing strategies, respectively, to match their relevant properties. Our conclusions highlight the importance of highly dynamic scheduling to optimize both the complexity and the success of load balancing. We also suggest the tuning of database schemata as a new anti-skew measure.

Data allocation is a key performance factor for parallel database systems (PDBS). This holds especially for data warehousing environments where huge amounts of data and complex analytical queries have to be dealt with. While there are several studies on data allocation for relational PDBS, the specific requirements of data warehouses have not yet been sufficiently addressed. In this study, we consider the allocation of relational data warehouses based on a star schema and utilizing bitmap index structures. We investigate how a multi-dimensional hierarchical data fragmentation of the fact table supports queries referencing different subsets of the schema dimensions. Our analysis is based on realistic parameters derived from a decision support benchmark. The performance implications of different allocation choices are evaluated by means of a detailed simulation model.

For complex queries in parallel database systems, substantial amounts of data must be redistributed between operators executed on different processing nodes. Frequently, such intermediate results cannot be held in main memory and must be stored on disk. To limit the ensuing performance penalty, a data allocation must be found that supports parallel I/O to the greatest possible extent. In this paper, we propose declustering even self-contained units of temporary data processed in a single operation (such as individual buckets of parallel hash joins) across multiple disks. Using a suitable analytical model, we find that the improvement of parallel I/O outweighs the penalty of increased fragmentation.

Skew effects are still a significant problem for efficient query processing in parallel database systems. Especially in shared-nothing environments, this problem is aggravated by the substantial cost of data redistribution. Shared-disk systems, on the other hand, promise much higher flexibility in the distribution of workload among processing nodes because all input data can be accessed by any node at equal cost. In order to verify this potential for dynamic load balancing, we have devised a new technique for skew-tolerant join processing. In contrast to conventional solutions, our algorithm is not restricted to estimating processing costs in advance and assigning tasks to nodes accordingly. Instead, it monitors the actual progression of work and dynamically allocates tasks to processors, thus capitalizing on the uniform access pathlength in shared-disk architectures. This approach has the potential to alleviate not only any kind of data-inherent skew, but also execution skew caused by query- external workloads, by disk contention, or simply by inaccurate estimates used in predictive scheduling. We employ a detailed simulation system to evaluate the new algorithm under different types and degrees of skew.

In parallelen Datenbanken, die für Decision-Support-Aufgaben wie z. B. Data Warehousing eingesetzt werden, spielen hohe Durchsatzraten, kurze Antwortzeiten und damit auch Lastbalancierungsfragen eine entscheidende Rolle. Dies gilt insbesondere für komplexe Operationen wie den relationalen Join. Das größte Problem bei seiner parallelen Ausführung sind nichtuniforme Daten- und Werteverteilungen (Skew), die nur begrenzt vorhersehbar sind und somit zur Laufzeit behandelt werden müssen. Dies ist in den verbreiteten Shared-Nothing-Rechnerarchitekturen jedoch nur schwer zu realisieren, da Datenumverteilungen mit hohem Zusatzaufwand verbunden sind. Wir schlagen daher ein dynamisches Lastbalancierungsverfahren auf Basis einer Shared-Disk-Architektur vor, welches aufgrund der uniformen Zugriffsstruktur weitaus effizienter arbeitet, als dies in Shared-Nothing-Systemen möglich ist. In einer Simulationsstudiezeigt es sich einem herkömmlichen prädiktiven Algorithmus deutlich überlegen.