Introduction
As mobile computing evolved, the diversity of client devices—ranging from high-performance laptops to resource-constrained personal digital assistants (PDAs)—created a significant challenge for service providers. Delivering a consistent user experience across these platforms requires dynamic content adaptation, such as transcoding video streams or resizing images at the network edge. However, these adaptations consume significant computational resources, including CPU and memory, which are not captured by traditional billing models. This paper proposes a middleware solution for "pervasive accounting," a system designed to track and charge for the specific computational costs associated with adaptive Quality of Service (QoS) in mobile environments.
The Challenge of Adaptive Services
Traditional accounting in wireless networks typically relies on connection duration or data volume. While sufficient for static content delivery, these metrics fail to account for the "tailoring costs" incurred at Service Access Points (SAPs). When a SAP downscales a high-definition video into a low-resolution format for a mobile user, it performs intensive processing that benefits the user but costs the provider. Without a mechanism to quantify and bill for these edge-node resources, providing high-quality adaptive services becomes economically unsustainable.
Furthermore, user mobility introduces technical hurdles. As users roam between different network domains, maintaining a continuous accounting session is difficult. Network instability and temporary disconnections (partitioning) can lead to lost metering data if the system relies on a centralized, always-connected billing server.
The SOMA-based Middleware Architecture
To address these issues, the authors developed a middleware platform built on the Secure and Open Mobile Agent (SOMA) framework. The core of this architecture is the use of Mobile Agents (MAs)—autonomous software entities capable of migrating their execution state across network nodes. This allows the accounting logic to follow the user physically and logically.
The architecture consists of four primary components:
- Metadata Service (MS): This component manages profiles using W3C standards like CC/PP (Composite Capability/Preference Profiles) and RDF (Resource Description Framework). It stores data regarding user billing tiers, device hardware limitations, and service pricing models.
- Monitoring Module (MM): A localized component residing on every node that tracks resource consumption. It is designed to be granular, distinguishing between system-level usage and the specific resources consumed by individual application-level adaptation tasks.
- Accounting Coordination Service (CoMAs): These agents serve as the decision-makers. They retrieve user and device profiles to determine the necessary level of service downscaling and instruct the Monitoring Module on which specific resources to track.
- Charging Service (CMAs): These agents act as "Care-of" entities for the user. They calculate an Estimate of Charge (EoC) before service delivery and generate final billing reports locally at the edge, minimizing the need for constant communication with a central server.
Key Innovations and Insights
Locality and Resilience
By processing accounting data at the edge where the service is consumed, the system remains functional even during network partitioning. Mobile Agents ensure that metering data is collected and processed locally, preventing the bottlenecks and data loss associated with centralized architectures.
Session Continuity via Mobile Agents
The "Care-of" entity model ensures that the accounting state is preserved during handoffs. When a user moves from one network domain to another, their dedicated Charging Agent migrates with them, maintaining the integrity of the billing session without requiring a complete restart of the accounting process.
Accounting as a Management Tool
The authors argue that pervasive accounting serves a dual purpose. Beyond simple billing, it acts as a network management tool. By charging for specific resource consumption, the system discourages "greedy" behavior, incentivizing users to select QoS levels that are appropriate for their devices and current network conditions, thereby reducing overall congestion.
Functional Validation: The Downtown Visitor Assistant
The system was evaluated through a functional case study called the Downtown Visitor Assistant (DVA). This scenario simulated tourists using PDAs to access multimedia guides via Wi-Fi access points in a city environment. The study demonstrated that the middleware could successfully:
- Provide real-time Estimates of Charge (EoC) based on the user's subscription tier (Gold, Silver, or Bronze).
- Dynamically adapt Video-on-Demand (VoD) flows while accurately tracking the CPU and memory overhead of the transcoding process.
- Apply different pricing logic based on time-of-day and user profiles.
- Minimize accounting overhead by selectively monitoring only the resources relevant to the specific service adaptation.
Conclusion and Future Work
The proposed middleware demonstrates the feasibility of using Mobile Agents to manage the complexities of accounting in pervasive, mobile environments. By shifting the focus from simple data volume to comprehensive resource consumption, the system enables sustainable delivery of adaptive QoS.
Future research directions include enhancing security to ensure non-repudiation (preventing users from denying usage), implementing decentralized micro-payment systems, and extending the model to peer-to-peer (P2P) ad-hoc networks where fixed infrastructure is unavailable. Addressing these challenges will be critical for the widespread adoption of adaptive multimedia services in increasingly fragmented mobile landscapes.
Summary updated