XMach-1: A Multi-User Benchmark for XML Data Management
Erhard Rahm, Timo Böhme · 2002
The benchmark defines a database of XML documents and a set of operations covering important characteristics of XML processing and querying. Key features of XMach-1 are scalability, multi-user simulation and the evaluation of the entire data management system. It has been sucessfully implemented for a variety of native XML database systems and XML-enabled relational and object-relational DBMS. The benchmark is based on a web application in order to model a typical use case of a XML data management system. The system architecture consists of four parts: the XML database, application servers, loaders to populate the database and browser clients. The application servers run a web (HTTP) server and other middleware components to support processing of the XML documents and to interact with the backend database. The XML database contains both document-centric and data-centric XML documents. The largest part is document-centric consisting of semi-structured documents with larger text portions such as books or essays. These documents are synthetically produced by a parameterizable generator. In order to achieve close-to-reality results when storing and querying text contents, text is generated from the 10,000 most frequent English words, using a distribution corresponding to natural language text. The documents varies in size (2-100 kB) as well as in structure (flat and deep element hierarchy). The second part of the database is a data-centric directory containing the metadata of the other documents such as document URL, name, insert- and update time. All data in this document is stored in attributes (no mixed