Title: A workload-generic and web-centric benchmark model design for multi- and hyper-media databases

Authors: Jia-Lang Seng, Yen-Fang Lin

Addresses: Department of MIS, National Chengchi University, Wenshan 116, Taipei, Taiwan. ' KPMG Taiwan, Wenshan 116, Taipei, Taiwan

Abstract: Hypermedia Database Management Systems (HDBMS) are the core technique of web-based Hypermedia Information Systems (WHIS). WHIS centres on the Enterprise Information Systems (EIS) over the internet platform. HDBMS capability and performance determines the success and failure of WHIS. HDBMS of web pages and navigation functions must be tested and measured in order to assess and evaluate its capacity and undertaking. Benchmark is a vital tool to measure and evaluate HDBMS performance. However, current database benchmarks such as TPC benchmarks, SAP R/3 SD, PeopleSoft e-bill, Oracle AS-BM, Wisconsin, and AS3AP benchmarks are mainly relational benchmarks not best used and suitable to test HDBMS. They are targeted at a specific industry, certain applications, and pre-determined query set. HyperModel benchmark, an object-oriented workload, is the closest test to evaluate HDBMS. However, it also suffers a structural deficiency of domain dependency and application unrepresentativeness as those current and standard benchmarks. The structural deficiency means that benchmark results vary as the application changes and the domains deviate from the standards. We perceive the issue from the workload design and from the user requirements. We believe a more domain-independent and application-representative approach is needed. A more generic and open workload model design is required to represent and reflect different user requirements on HDBMS applications. In this study, we present a common carrier concept of workload requirements formulation. We perceive the workload model design as a process of requirements representation, transformation, and generation. The common carrier has to be developed from the generic models in ODMG, OHSWG, and TPC standards. A four-component model is created in this research to capture and compile the requirements. The components are the object model, application model, navigation model, and control model. Each component is implemented with a corresponding set of editors and generators. A system prototype is constructed to prove the feasibility and generalisation. Users can minimise their inputs and coding efforts by entering their requirements into the prototype. The implementation will generate different user-centric workloads based on the inputs. Workloads can determine performance profile. Hence, we secure a set of more realistic and reproducible test results from the model and the implementation.

Keywords: database benchmark; performance analysis; workload model; standards; services; web-based hypermedia information systems; hypermedia database management systems; object model; navigation model; application model; control model; internet; benchmarking; multimedia.

DOI: 10.1504/IJSS.2004.005686

International Journal of Services and Standards, 2004 Vol.1 No.1, pp.18 - 38

Published online: 18 Nov 2004 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article