Workload Model Based Dynamic Adaptation of Social Internet of Vehicles

Kazi Masudul Alam, Mukesh Kumar Saini, Abdulmotaleb El-SaddikSensors 2015 · OpenReview

TL;DR

This paper introduces a workload model for the Social Internet of Vehicles (SIoV) to understand and manage the immense data generated by connected vehicles. By analyzing subsystem workloads under various conditions, they propose dynamic adaptation strategies to ensure efficient operation and support the development of intelligent transport systems.

What Problem Does This Paper Address?

The increasing number of connected vehicles, forming the Social Internet of Vehicles (SIoV), generates a massive amount of data. This data is crucial for smart city applications, but the sheer volume and complexity make it challenging to test and deploy SIoV applications effectively. The paper addresses the need for a systematic way to model and understand the "workload" (storage and computational demands) of different components within the SIoV ecosystem (vehicles, infrastructure, homes, and the cloud). Without this understanding, it's difficult to design systems that can dynamically adapt to varying conditions and ensure reliable performance for intelligent transport systems.

Methods

The authors propose an incremental approach to model SIoV workloads. They start with a basic scenario involving a single vehicle and then extend it to more complex scenarios with multiple vehicles, infrastructures, and homes. The core of their method involves developing analytical models to quantify the data storage requirements for each subsystem. These models consider factors like vehicle speed, arrival rate, communication range, and data rates. To validate their models, they implement a simulator using SUMO for mobility traces and Java for virtual processes, allowing them to run extensive simulations under various parameter settings. Based on the insights gained from these models and simulations, they then propose dynamic adaptation strategies for each subsystem to manage workloads and optimize performance.

Datasets

Not discussed in the paper. The authors mention using OpenStreetMap for map area selection and SUMO for generating mobility traces, but these are tools for simulation rather than datasets for training or evaluation.

Key Insights and Ideas

  • SIoV as a Cyber-Physical System: The paper frames SIoV as a cloud-centric cyber-physical system, emphasizing the tight feedback loop between physical vehicle actions and cloud-based services.
  • Workload as a Key Metric: They highlight "workload" (primarily storage requirements) as a critical factor for understanding and managing SIoV performance, especially in the context of Big Data.
  • Incremental Modeling: The approach of building from a basic model to complex scenarios allows for a systematic understanding of how different factors influence workload.
  • Dynamic Adaptation is Crucial: The paper strongly advocates for dynamic adaptation strategies, suggesting that SIoV components need to be aware of their workload and adjust their behavior (e.g., data collection, communication range) to maintain performance.
  • Parameter Interdependencies: The simulations reveal clear relationships between various parameters (arrival rate, speed, communication range) and subsystem workloads, providing valuable insights for system design.

Key Results

The paper presents extensive simulation results demonstrating the impact of various parameters on subsystem storage requirements:

  • Arrival Rate (p): Storage requirements grow with increasing vehicle arrival rate (ψ ∝ ρ). This is attributed to larger platoon sizes and more interactions.
  • Vehicle Speed (v): Storage requirements decrease with increasing vehicle speed (ψ ∝ 1/v). Higher speeds reduce the connection life of vehicles, leading to fewer interactions and less data to store.
  • Communication Range (rv): Storage requirements increase with vehicle communication range (ψ ∝ rv). A larger range allows more vehicles to join platoons and exchange data.
  • Infrastructure Density (L): Storage requirements increase with the distance between infrastructures (ψ ∝ L). Vehicles travel longer distances, accumulating more data before reaching the next infrastructure.

The authors also propose adaptation strategies for vehicles, infrastructures, and homes, aiming to balance data collection with workload management. For instance, vehicles can curtail data collection if their workload exceeds a threshold, and infrastructures can divert traffic to reduce arrival rates.

Limitations and Future Work

  • Limitations:
    • The paper focuses primarily on data storage requirements as the measure of workload. Other computational demands are not as extensively detailed.
  • The simulation assumes error-free network connections, which might not reflect real-world scenarios.
  • The adaptation strategies are presented as examples and a comprehensive analysis is deferred to a future article.
  • Future Work:
    • Developing a robust cyber-physical IoV simulator that fully incorporates adaptation strategies.
  • Integrating real vehicles with the simulated environment to gain practical deployment insights.
  • Exploring the integration of heterogeneous IoT domains through the cyber layer.
  • Developing a multimedia sensory dataset for IoV applications, including OBD and smartphone data.

Summary updated

Abstract

Social Internet of Things (SIoT) has gained much interest among different research groups in recent times. As a key member of a smart city, the vehicular domain of SIoT (SIoV) is also undergoing steep development. In the SIoV, vehicles work as sensor-hub to capture surrounding information using the in-vehicle and Smartphone sensors and later publish them for the consumers. A cloud centric cyber-physical system better describes the SIoV model where physical sensing-actuation process affects the cloud based service sharing or computation in a feedback loop or vice versa. The cyber based social relationship abstraction enables distributed, easily navigable and scalable peer-to-peer communication among the SIoV subsystems. These cyber-physical interactions involve a huge amount of data and it is difficult to form a real instance of the system to test the feasibility of SIoV applications. In this paper, we propose an analytical model to measure the workloads of various subsystems involved in the SIoV process. We present the basic model which is further extended to incorporate complex scenarios. We provide extensive simulation results for different parameter settings of the SIoV system. The findings of the analyses are further used to design example adaptation strategies for the SIoV subsystems which would foster deployment of intelligent transport systems.
BibTeX
@inproceedings{alam2015workload,
  author = {Kazi Masudul Alam and Mukesh Kumar Saini and Abdulmotaleb El-Saddik},
  title = {{Workload Model Based Dynamic Adaptation of Social Internet of Vehicles}},
  booktitle = {Sensors},
  year = {2015},
  url = {https://openreview.net/forum?id=pqlQ1-w4KtU},
}

Loading related content...

Figure