[Paper Review] Analysis of Content-Oriented Heterogeneous Networks with D2D and Cognitive Communications

S. Sinem Kafıloğlu, Gürkan Gür+1 more|arXiv (Cornell University)|Aug 2, 2018

Caching and Content Delivery18 references4 citations

TL;DR

This paper proposes a content-aware caching and cognitive radio-enabled satellite-terrestrial network with device-to-device (D2D) overlaying, modeled via a continuous-time Markov chain (CTMC) to optimize energy efficiency (EE) and goodput. It demonstrates that popularity-aware caching outperforms LRU and FIFO in goodput under high request loads, while D2D overlaying enhances both EE and system capacity.

ABSTRACT

Content-oriented operation with D2D communications and spectrum sharing in the form of cognitive communications are expected to be enablers for cost-efficient nextgeneration wireless systems with bandwidth-hungry services. To this end, caching is an important building block of these systems. In this work, we model such a heterogeneous network and model a popularity-aware caching algorithm where we try to keep more popular contents in caches of system units with a higher probability compared to less popular contents. Moreover, for improving the usage of the channel capacity, overlaying for device-to-device (D2D) transmissions (unless these transmission interfere with each other) is enabled. We model our system with an analytical continuous time Markov chain (CTMC). We have a heterogeneous network architecture consisting of a LEO satellite, base station (BS) and a set of devices within the boundary of BS coverage that can get service from the satellite, from the BS or in D2D mode. Our devices operate in primary mode when they get service from the satellite but in secondary mode when they get service from the BS or some device in D2D mode over the terrestrial link. We evaluate our system for energy efficiency (EE) and goodput metrics. We looked how popularity-aware caching, integration of universal source concept and overlaying in D2D mode affect the network performance. Besides, we investigate how resource allocation (RA) and performance metrics are tangled.

Motivation & Objective

• To model a heterogeneous satellite-terrestrial network with D2D and cognitive communications for improved energy efficiency and goodput.
• To design and evaluate a popularity-aware caching algorithm that prioritizes frequently requested content in system caches.
• To integrate a universal source concept to ensure content availability beyond local caches.
• To enable and evaluate overlaying D2D communications that reuse spectrum without interference to boost network capacity.
• To analyze the trade-offs between resource allocation, mode selection, and performance metrics like EE and goodput.

Proposed method

• Models the network as a continuous-time Markov chain (CTMC) with states representing user modes (satellite, BS, D2D) and content availability.
• Defines state transitions for PU (satellite) arrivals and departures, D2D mode operations, and HU (hybrid user) content requests.
• Introduces a mode selection mechanism based on cache availability, channel state, and weighted mode preferences.
• Applies a popularity-aware caching algorithm that increases the probability of caching frequently requested content.
• Enables spectrum overlaying in D2D mode to allow simultaneous transmissions without interference, improving spectral efficiency.
• Integrates a universal source to provide content not available in local caches, ensuring end-to-end content delivery.

Experimental results

Research questions

• RQ1 How does popularity-aware caching affect energy efficiency and goodput compared to traditional caching policies like LRU and FIFO?
• RQ2 To what extent does enabling overlaying D2D communication improve network capacity and energy efficiency?
• RQ3 How does the integration of a universal source impact system performance under varying request loads?
• RQ4 What is the effect of user mode selection weights (satellite, BS, D2D) on EE and goodput?
• RQ5 How do PU (satellite) arrival rates on the terrestrial link influence system performance metrics?

Key findings

• Popularity-aware caching outperforms LRU and FIFO in overall system goodput, especially under high user request rates (λHU), achieving up to 46.04 Mbps goodput.
• D2D overlaying improves both energy efficiency and goodput, with goodput increasing by up to 48.5 Mbps as D2D mode weight (rdev) increases.
• Energy efficiency (EPBHU) decreases with higher D2D usage due to lower transmission power and shorter durations, with values dropping to 0.16 J/packet at rdev = 0.9.
• The universal source integration does not degrade goodput and enables access to off-network content, though it has minimal impact on EE under light load.
• High PU arrival rates on the terrestrial link reduce goodput but do not significantly affect EE, indicating robustness in interference-limited scenarios.
• Turning off D2D mode (rdev = 0) degrades both EE and goodput, while setting BS mode weight to zero while increasing D2D weight reduces goodput due to underutilization of BS cache.