Abstract
When Cloud Platform for Computer Aided Engineering Simulation (CPCAES) are developed to use Service-Oriented Architecture (SOA), it is necessary to accurately express user service requirements and the semantics of existing services to perform effective ontology matching. Existing algorithms like OWL-S/UDDI often lack the precision required for complex CAE applications. This paper presents a specialized service requirement ontology for CAE simulation, encompassing resource, domain, and job requirements. We define mapping relationships between this ontology and OWL-S, establish matching decision rules, and propose a similarity-based matching algorithm. Evaluation against traditional methods shows our approach achieves approximately 90% precision and recall, offering a more suitable solution for CPCAES.
Introduction
Computer-Aided Engineering (CAE) simulations, such as stress analysis and fluid dynamics, are computationally intensive tasks. While Cloud Platforms for CAE (CPCAES) aim to simplify access to these tools, matching specific user requirements—such as software versions, core counts, and licenses—to available cloud services remains a significant challenge. Standard protocols like OWL-S and UDDI are often too generic, failing to capture the technical nuances of high-performance engineering simulations, which leads to low matching precision.
The CAE Service Requirement Ontology
To address these limitations, we developed a specialized ontology that extends the standard OWL-S ServiceProfile. This RequirementProfile captures CAE-specific data across several dimensions:
- RequirementResource: Software licenses, CPU configurations, and memory.
- RequirementDomain: The specific engineering field (e.g., heat transfer, collision analysis).
- RequirementJob: Operational constraints like priority, parallel scale, and response time.
Ontology Mapping and Similarity Algorithm
A critical component of our system is the formal mapping between the custom RequirementProfile and the standard ServiceProfile. This ensures compatibility with broader semantic web standards while allowing for specialized matching.
Our matching algorithm moves beyond binary "yes/no" results. It calculates a numerical similarity score (ranging from 0 to 1) based on the semantic distance between nodes in the ontology tree. The final match is a weighted sum of four primary factors:
- Input matching
- Output matching
- Precondition matching (Resource and Domain requirements)
- Quality of Service (QoS) matching (Job requirements)
This weighted approach allows the system to prioritize certain parameters, such as specific software versions over memory size, depending on user needs.
Experimental Evaluation
The algorithm was evaluated using 100 services from a CPCAES prototype. We compared our approach against the classic OWL-S/UDDI matching algorithm and a general semantic similarity algorithm.
Key results include:
- High Precision and Recall: Both metrics reached approximately 90%, significantly outperforming the OWL-S/UDDI baseline, especially as requirement complexity increased.
- Qualitative Accuracy: The system successfully distinguished between similar services (e.g., different versions of ANSYS for different analysis types) that generic matchers often confused.
- Real-world Application: The algorithm is currently integrated into the second prototype of the National Supercomputing Center in Changsha’s cloud platform.
Conclusion and Future Work
The proposed semantic matching algorithm provides a robust framework for service discovery in CAE cloud environments. While currently optimized for CAE, future work will explore adapting the ontology for other domains like bioinformatics. Additionally, we aim to transition from simple matching to fully automatic service composition using Petri nets and the Activiti process engine to manage complex simulation workflows.
Summary updated