Search results

The purpose of this paper is to propose general parallelism techniques for holistic twig join algorithms to process queries against Extensible Markup Language (XML) databases on a multi‐core system.

The parallelism techniques comprised data and task parallelism. As for data parallelism, the paper adopted the stream‐based partitioning for XML to partition XML data as the basis of parallelism on multiple CPU cores. The XML data partitioning was performed in two levels. The first level was to create buckets for creating data independence and balancing loads among CPU cores; each bucket was assigned onto a CPU core. Within each bucket, the second level of XML data partitioning was performed to create finer partitions for providing finer parallelism. Each CPU core performed the holistic twig join algorithm on each finer partition of its own in parallel with other CPU cores. In task parallelism, the holistic twig join algorithm was decomposed into two main tasks, which were pipelined to create parallelism. The first task adopted the data parallelism technique and their outputs were transferred to the second task periodically. Since data transfers incurred overheads, the size of each data transfer needed to be estimated cautiously for achieving optimal performance.

The data and task parallelism techniques contribute to good performance especially for queries having complex structures and/or higher values of query selectivity. The performance of data parallelism can be further improved by task parallelism. Significant performance improvement is attained by queries having higher selectivity because more outputs computed by the second task is performed in parallel with the first task.

The proposed parallelism techniques primarily deals with executing a single long‐running query for intra‐query parallelism, partitioning XML data on‐the‐fly, and allocating partitions on CPU cores statically. During the parallel execution, presumably there are no such dynamic XML data updates.

The effectiveness of the proposed parallel holistic twig joins relies fundamentally on some system parameter values that can be obtained from a benchmark of the system platform.

The paper proposes novel techniques to increase parallelism by combining techniques of data and task parallelism for achieving high performance. To the best of the author's knowledge, this is the first paper of parallelizing the holistic twig join algorithms on a multi‐core system.

The purpose of this paper is to propose Extensible Markup Language (XML) data partitioning schemes that can cope with static and dynamic allocation for parallel holistic twig joins: grid metadata model for XML (GMX) and streams‐based partitioning method for XML (SPX).

GMX exploits the relationships between XML documents and query patterns to perform workload‐aware partitioning of XML data. Specifically, the paper constructs a two‐dimensional model with a document dimension and a query dimension in which each object in a dimension is composed from XML metadata related to the dimension. GMX provides a set of XML data partitioning methods that include document clustering, query clustering, document‐based refinement, query‐based refinement, and query‐path refinement, thereby enabling XML data partitioning based on the static information of XML metadata. In contrast, SPX explores the structural relationships of query elements and a range‐containment property of XML streams to generate partitions and allocate them to cluster nodes on‐the‐fly.

GMX provides several salient features: a set of partition granularities that balance workloads of query processing costs among cluster nodes statically; inter‐query parallelism as well as intra‐query parallelism at multiple extents; and better parallel query performance when all estimated queries are executed simultaneously to meet their probability of query occurrences in the system. SPX also offers the following features: minimal computation time to generate partitions; balancing skewed workloads dynamically on the system; producing higher intra‐query parallelism; and gaining better parallel query performance.

The current status of the proposed XML data partitioning schemes does not take into account XML data updates, e.g. new XML documents and query pattern changes submitted by users on the system.

Note that effectiveness of the XML data partitioning schemes mainly relies on the accuracy of the cost model to estimate query processing costs. The cost model must be adjusted to reflect characteristics of a system platform used in the implementation.

This paper proposes novel schemes of conducting XML data partitioning to achieve both static and dynamic workload balance.