[Paper Review] Performance Comparison of Dual Connectivity and Hard Handover for LTE-5G Tight Integration

Michele Polese, Marco Mezzavilla+1 more|arXiv (Cornell University)|Jul 19, 2016

Advanced MIMO Systems Optimization21 citations

TL;DR

This paper evaluates dual connectivity (DC) and hard handover (HH) for tight integration between LTE and mmWave 5G networks, using a custom ns-3 simulator with 5G mmWave and LTE modules. DC achieves 106.70 Mbit/s average PDCP throughput and 5.1 ms latency, outperforming HH's 104.98 Mbit/s and 18.1 ms due to faster switching and reduced buffering during link changes.

ABSTRACT

Communications at frequencies above 10 GHz (the mmWave band) are expected to play a major role for the next generation of cellular networks (5G), because of the potential multi-gigabit, ultra-low latency performance of this technology. mmWave frequencies however suffer from very high isotropic pathloss, which may result in cells with a much smaller coverage area than current LTE macrocells. High directionality techniques will be used to improve signal quality and extend coverage area, along with a high density deployment of mmWave base stations (BS). However, when propagation conditions are hard and it is difficult to provide high quality coverage with mmWave BS, it is necessary to rely on previous generation LTE base stations, which make use of lower frequencies (900 MHz - 3.5 GHz), which are less sensitive to blockage and experience lower pathloss. In order to provide ultra-reliable services to mobile users there is a need for network architectures that tightly and seamlessly integrate the LTE and mmWave Radio Access Technologies. In this paper we will present two possible alternatives for this integration and show how simulation tools can be used to assess and compare their performance.

Motivation & Objective

• To evaluate and compare the performance of dual connectivity (DC) and hard handover (HH) in tightly integrated LTE and mmWave 5G networks.
• To address the challenge of maintaining ultra-reliable, low-latency connectivity in mmWave bands, which suffer from high pathloss and blockage.
• To develop and validate a simulation framework in ns-3 that integrates 5G mmWave and LTE protocol stacks for accurate performance assessment.
• To quantify the impact of mobility speed and X2 interface latency on signaling overhead and user plane performance.

Proposed method

• Implemented a hybrid ns-3 simulator integrating the NYU 5G mmWave protocol stack with the standard LTE module in ns-3.
• Enabled dual connectivity by allowing a user equipment (UE) to simultaneously connect to both an LTE eNB and a mmWave eNB, with fast switching via a single RRC message.
• Used a PDCP-layer integration point to enable seamless handover-like switching between RATs without full re-establishment.
• Simulated a realistic urban scenario with 28 GHz mmWave eNBs, 2.1 GHz LTE eNBs, and UE mobility at 2 m/s along a path with blockages.
• Configured identical random number seeds across DC and HH simulations to ensure comparable channel conditions for fair comparison.
• Measured key metrics including PDCP throughput, latency, RLC buffer occupancy, and RRC signaling load under varying X2 latency and UE speed conditions.

Experimental results

Research questions

• RQ1 How does dual connectivity compare to hard handover in terms of end-to-end latency and throughput during mmWave link outages?
• RQ2 What is the impact of X2 interface latency on RRC signaling overhead in dual connectivity and hard handover scenarios?
• RQ3 How does UE mobility speed affect the performance of dual connectivity versus hard handover in mmWave coverage holes?
• RQ4 To what extent does dual connectivity reduce packet loss and buffering compared to hard handover during RAT switching?
• RQ5 Can dual connectivity achieve lower latency and higher reliability than hard handover in mmWave-based 5G networks?

Key findings

• Dual connectivity achieved an average PDCP-layer throughput of 106.70 Mbit/s, significantly higher than hard handover’s 104.98 Mbit/s.
• Dual connectivity reduced average latency to 5.1 ms, compared to 18.1 ms for hard handover, due to faster switching and minimal buffering.
• During outage events, dual connectivity maintained latency below 40 ms, while hard handover exhibited a spike of 287 ms.
• Dual connectivity reduced RRC signaling overhead per user, enabling scalability under limited control plane capacity.
• The performance gain of dual connectivity was consistent across different UE speeds (2–16 m/s) and X2 latencies (0.1–10 ms).
• The simulation framework enabled fair, repeatable comparison between DC and HH by synchronizing channel variations via identical random seeds.