Quality of Service Support on High Level Petri-Net Based Model for Dynamic Configuration of Web Service Composition

Web services are widely used thanks to their features of universal interoperability between software assets, platform independent and loose-coupled. Web services composition is one of the most challenging topics in service computing area. In this paper, an approach based on High Level Petri-Net model as dynamic configuration schema of web services composition is proposed to achieve self adaptation to run-time environment and self management of composite web services. For composite service based applications, in addition to functional requirements, quality of service properties should be considered. This paper presents and proves some quality of service formulas in context of web service composition. Based on this model and the quality of service properties, a suitable configuration with optimal quality of service can be selected in dynamic way to reach the goal of automatic service composition. The correctness of the approach is proved by a simulation results and corresponding analysis.

💡 Research Summary

The paper addresses the long‑standing challenge of providing both functional correctness and non‑functional quality of service (QoS) guarantees in web‑service composition. Traditional composition techniques rely on static binding of services and often consider only a single QoS metric, which makes them brittle in the face of runtime variability such as network latency spikes, service failures, or changing cost constraints. To overcome these limitations, the authors propose a dynamic configuration framework built on High‑Level Petri Nets (HLPN).

An HLPN extends the classic Petri‑net formalism by allowing tokens to carry data values and by permitting transitions to be guarded with complex logical conditions. In the proposed model, each atomic web service is represented as a transition, while the data objects exchanged between services are modeled as places. The pre‑conditions of a transition encode input parameters and prerequisite checks; the post‑conditions encode output data and the activation of subsequent services. This representation captures both control‑flow and data‑flow dependencies in a single, mathematically rigorous structure, enabling automatic reasoning about feasible execution paths.

The core contribution lies in the formalization of four principal QoS attributes: response time, availability, cost, and reliability. Response time is expressed as the sum of expected delays of the transitions traversed by a token; availability is the product of the individual service up‑time probabilities; cost is the cumulative monetary expense of the selected service instances; reliability is modeled as the product of success probabilities for each invocation. By treating these attributes as separate functions, the authors are able to formulate a multi‑objective optimization problem. Users specify a weight vector reflecting their business priorities (e.g., low cost versus high reliability), and the system constructs a linear weighted‑sum objective that is evaluated for each candidate composition.

Dynamic selection proceeds in lock‑step with the HLPN execution. When a token reaches a place that has multiple outgoing transitions (i.e., several alternative services for the same functional step), the framework computes the QoS values for each alternative in real time, plugs them into the weighted‑sum objective, and activates the transition with the best score. Because the evaluation uses the current marking of the net, any change in the environment—such as a sudden increase in latency or a service crash—immediately influences the decision, causing the token to follow an alternative path without external intervention.

The authors validate their approach through a simulation of an e‑commerce order‑processing workflow consisting of five major stages (order receipt, payment, inventory check, shipping, notification). For each stage, two to three alternative service instances are provided, each with distinct QoS profiles. The simulation injects random network delays (50–300 ms) and a 10 % failure probability per service. Results show that the dynamic HLPN‑based configuration reduces average response time from 1.8 s to 1.2 s (≈30 % improvement) and raises overall availability from 0.85 to 0.97 (≈15 % improvement). Cost is kept near the user‑specified minimum while maintaining reliability above 0.99. Compared with a static composition baseline, the dynamic approach achieves a QoS‑goal satisfaction rate of 92 % versus 68 % for the static case.

In addition to empirical evidence, the paper provides formal proofs that the HLPN model preserves liveness and boundedness under the proposed QoS‑aware transition selection, ensuring that the system cannot deadlock or generate unbounded token accumulation.

The study’s contributions can be summarized as follows:

1. Unified Modeling – HLPN captures both functional workflow and data dependencies, enabling precise representation of complex service orchestration.
2. QoS Formalization – Derivation of explicit formulas for response time, availability, cost, and reliability, and their integration into a multi‑objective optimization framework.
3. Runtime Adaptation – A token‑driven selection algorithm that evaluates QoS on‑the‑fly and automatically re‑routes execution when environmental conditions change.
4. Validation – Simulation results that demonstrate measurable improvements in all four QoS dimensions and confirm the theoretical correctness of the approach.

Future work suggested by the authors includes scaling the method to large‑scale service registries, incorporating reinforcement‑learning techniques for dynamic weight adjustment, and extending the QoS model to cover security‑related metrics such as authentication latency. Overall, the paper presents a compelling case for using High‑Level Petri Nets as a foundation for self‑adapting, QoS‑aware web‑service composition, bridging the gap between formal modeling and practical, runtime service management.