OSCAR: A Collaborative Bandwidth Aggregation System

Karim Habak, Khaled A. Harras, Moustafa YoussefarXiv 2014 · OpenReview

TL;DR

OSCAR is a novel system that aggregates bandwidth from multiple network interfaces on mobile devices, both individually and collaboratively with nearby devices. It achieves significant throughput gains (up to 150%) without requiring changes to existing infrastructure, while also offering users control over cost and energy consumption.

What Problem Does This Paper Address?

The paper tackles the growing challenge of mobile data demand, driven by user expectations for constant connectivity and the proliferation of multi-interface devices. Current mobile operating systems typically allow users to utilize only one network interface at a time, creating a bottleneck. While existing solutions for bandwidth aggregation exist, they often suffer from a high deployment barrier, requiring modifications to legacy servers or applications, or focusing solely on maximizing throughput without considering energy or cost efficiency. The authors aim to provide a deployable, collaborative, and multi-objective solution.

Methods

OSCAR introduces a multi-objective, incentive-based, collaborative, and deployable bandwidth aggregation system. Its core components include:

  • Context Awareness Manager: Gathers information about user requirements, application traffic, network interfaces, battery status, and nearby OSCAR-enabled devices.
  • Backward Compatibility Manager: Ensures OSCAR works with existing servers and applications without modifications.
  • Packet-Oriented Scheduling Manager: Exploits servers that support connection resumption or are OSCAR-enabled to achieve finer-grained scheduling.
  • Incentives Manager: Implements a credit-based system using a virtual bank to encourage users to share their bandwidth.
  • OSCAR Scheduler: A multi-objective scheduler that can be configured to prioritize throughput maximization, energy minimization, or cost minimization, subject to user-defined constraints.
  • Communication Protocols: Developed for efficient and secure communication between OSCAR clients, neighbors, and OSCAR-enabled servers.

A key innovation is the ability to seamlessly switch between connection-oriented (for legacy servers) and packet-oriented (for OSCAR-enabled or resume-supporting servers) modes, and to leverage neighboring devices' underutilized bandwidth through an incentive mechanism.

Datasets

Not discussed in the paper. The evaluation relies on experimental setups and simulations using defined parameters for network interfaces, traffic, and application characteristics.

Key Insights and Ideas

  • Deployability is Key: The paper emphasizes that solutions must work without modifying existing infrastructure or applications to gain widespread adoption. OSCAR's design prioritizes this.
  • Beyond Throughput: While maximizing throughput is important, users also care about energy consumption and financial cost. OSCAR's multi-objective scheduler allows users to balance these trade-offs.
  • Collaboration is a Win-Win: By incentivizing users to share their bandwidth, OSCAR creates a scenario where users can access more bandwidth when needed and earn credits for sharing, benefiting both parties.
  • Leveraging Legacy Features: The ability to utilize connection resume functionality on existing servers significantly boosts performance, even without full OSCAR integration.
  • Context is Crucial: OSCAR's effectiveness hinges on its ability to understand the device's context (user needs, network conditions, battery) to make optimal scheduling decisions.

Key Results

  • Significant Throughput Gains: In throughput maximization mode, OSCAR provides up to a 150% enhancement in throughput compared to current operating systems, even when interacting with legacy servers.
  • Performance Boost with OSCAR/Resume-Enabled Servers: This performance gain further increases when connecting to OSCAR-enabled or resume-supporting servers, approaching the theoretical upper bound.
  • Cost and Energy Control: OSCAR allows users to set limits on average cost and energy consumption per unit data, keeping them within acceptable ranges.
  • Effective Neighbor Sharing: The system effectively leverages bandwidth from neighboring devices, with performance saturating around a 70% neighbor sharing ratio due to local interface limitations.
  • Interface Heterogeneity Handling: OSCAR outperforms simpler schedulers like Round Robin and Weighted Round Robin when dealing with heterogeneous network interfaces, by considering application characteristics and resume support.

Limitations and Future Work

  • Trust Assumption: OSCAR assumes that all OSCAR components are trusted, and security aspects of the software are outside the scope of this paper.
  • Infeasible Solutions: The paper acknowledges that user-defined constraints might lead to infeasible solutions, proposing an iterative selection policy to guide users towards feasible configurations.
  • Future Work:
    • Extending the system to optimize for multiple objectives, including trust, in addition to throughput, cost, and energy.
  • Implementing OSCAR on mobile operating systems.
  • Developing adaptive scheduling strategies that are sensitive to user profiles and real-time needs.

Summary updated

Abstract

The exponential increase in mobile data demand, coupled with growing user expectation to be connected in all places at all times, have introduced novel challenges for researchers to address. Fortunately, the wide spread deployment of various network technologies and the increased adoption of multi-interface enabled devices have enabled researchers to develop solutions for those challenges. Such solutions aim to exploit available interfaces on such devices in both solitary and collaborative forms. These solutions, however, have faced a steep deployment barrier. In this paper, we present OSCAR, a multi-objective, incentive-based, collaborative, and deployable bandwidth aggregation system. We present the OSCAR architecture that does not introduce any intermediate hardware nor require changes to current applications or legacy servers. The OSCAR architecture is designed to automatically estimate the system's context, dynamically schedule various connections and/or packets to different interfaces, be backwards compatible with the current Internet architecture, and provide the user with incentives for collaboration. We also formulate the OSCAR scheduler as a multi-objective, multi-modal scheduler that maximizes system throughput while minimizing energy consumption or financial cost. We evaluate OSCAR via implementation on Linux, as well as via simulation, and compare our results to the current optimal achievable throughput, cost, and energy consumption. Our evaluation shows that, in the throughput maximization mode, we provide up to 150% enhancement in throughput compared to current operating systems, without any changes to legacy servers. Moreover, this performance gain further increases with the availability of connection resume-supporting, or OSCAR-enabled servers, reaching the maximum achievable upper-bound throughput.
BibTeX
@misc{habak2014oscar,
  author = {Karim Habak and Khaled A. Harras and Moustafa Youssef},
  title = {{OSCAR: A Collaborative Bandwidth Aggregation System}},
  year = {2014},
  eprint = {1401.1258},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1401.1258},
}

Loading related content...

Figure