Distributed mobility management based on flat network architecture

Yu Liu, Zhijun Zhao, Tao Lin, Tang Hui2010 · DOI

Abstract

Mobility management is crucial for mobile networks to support certain quality of service (QoS) requirements. In this paper, we developed a flat network architecture rather than traditional two-layer mobility management architecture to reduce network elements and lighten networking load. Based on this architecture, we propose a Distributed Mobility Management (DMM) scheme that separates a mobile node's identity from its location and employs localized bicasting to minimize handoff latency and signaling overhead.

Introduction

In traditional mobile networking, such as Mobile IP, managing a device as it moves between access points is a resource-intensive operation. These systems typically rely on a two-layer hierarchy consisting of a "Home Agent" (HA), which acts as the permanent anchor for a device, and a "Foreign Agent" (FA), which manages the device in its current visiting network.

This centralized structure introduces several significant challenges:

  1. High Signaling Cost: Constant updates between the HA and FA consume substantial network resources.
  2. Handoff Latency: Transitions between base stations often result in connectivity gaps while the network reconfigures, leading to packet loss or jitter.
  3. Scalability Bottlenecks: Centralized agents like the HA become single points of failure and performance bottlenecks as the number of mobile users grows.

To address these issues, this paper proposes "flattening" the network architecture to create a more efficient, decentralized mobility management system.

Proposed Flat Architecture

The core of the proposed system is the replacement of the HA/FA hierarchy with a single tier of Access Gateways (AGWs). In this flat architecture, every AGW manages its own local domain and communicates directly with other gateways, eliminating the need for a centralized anchor.

The proposed flat network architecture featuring distributed Access Gateways (AGWs) connected in a decentralized manner.

Location and Identity Separation

A key innovation in this DMM scheme is the separation of a Mobile Node’s (MN) Identity (ID) from its Location (IP). While a device's ID remains constant globally, its IP address changes dynamically as it moves between AGW domains. To track these changes without a central server, the system utilizes a Distributed Hash Table (DHT). AGWs store (ID, IP) pairs in the DHT, allowing any node to locate another through a decentralized lookup process.

Handoff Mechanism and Local Bicasting

To achieve seamless transitions, the paper introduces a localized bicasting mechanism. When a handoff is initiated, the "serving" AGW replicates incoming data packets and sends them to both the current location and the "target" AGW simultaneously.

Sequence diagram of the handoff procedure, illustrating the preparation and execution phases including bicasting.

By performing bicasting at the network edge (the AGW) rather than at the data source (the Correspondent Node), the system achieves seamless handoff with significantly lower delivery costs. The handoff process is divided into two phases:

  • Handoff Preparation Phase: The network determines the target AGW, transfers context, and initiates bicasting.
  • Handoff Execution Phase: The link-layer switch occurs, and the (ID, IP) pair is updated in the DHT.

Performance Analysis

The authors evaluated the DMM approach using analytical modeling and numerical simulations, comparing it against Fast Mobile IP (FMIP). Key findings include:

  • Signaling Efficiency: The DMM architecture consistently outperforms FMIP in signaling overhead, particularly at low Call-to-Mobility Ratios (CMR), where users move frequently relative to data reception.
  • Latency Reduction: The method successfully masks movement detection and address configuration phases. Under ideal conditions, handoff latency is reduced to the theoretical minimum: the time required for the link-layer switch ($t_{l2switch}$).
  • Zero Packet Loss: Mathematical frameworks indicate that packet loss can be eliminated if AGWs are equipped with sufficient buffers ($C_b$) to hold data during the brief link-layer transition.

Conclusion and Future Work

By flattening the network and decentralizing mobility management, this architecture offers a scalable solution for modern mobile demands. However, several areas remain for future exploration, including the specific latency overhead of DHT lookups in global-scale networks and the development of dynamic buffer management algorithms. Future research will focus on implementing this model in heterogeneous environments, such as mixed 5G and Wi-Fi networks.

Summary updated

Abstract

Mobility management is crucial for mobile networks to support certain quality of service (QoS) requirements. In this paper, we developed a flat network architecture rather than traditional two-layer mobility management architecture to reduce network elements and lighten networking load. Based on the
BibTeX
@misc{liu2010distributed,
  author = {Yu Liu and Zhijun Zhao and Tao Lin and Tang Hui},
  title = {{Distributed mobility management based on flat network architecture}},
  year = {2010},
  doi = {10.4108/icst.wicon2010.8728},
}
Figure