This page catalogs research projects that make use of Aether, with the goal of illustrating how Aether can be used in the lab to investigate and demonstrate new features. If you have a project that uses Aether, please add a summary/abstract of your work (along with links to any results or artifacts) to this page.
Tegra: A Fast, Flexible, and Scalable Cloud-Native 5G Core
To support the rapidly evolving mobile use cases (e.g., AR/VR, autonomous driving, and massive IoT), the 5G mobile core (5GC) is being architected as a service-based architecture (SBA) workload running on private/public clouds. Yet, the current proposals to improve its performance still revert to old methods used in traditional NFV-based mobile core designs (e.g., consolidating functions on dedicated servers). It is critical to understand whether there is a fundamental tradeoff between achieving greater flexibility and higher performance in SBA-based mobile cores.
To answer this question, we conduct an in-depth study of a 5G-compliant open-source mobile core (i.e., Aether) to characterize its various bottlenecks. Our measurements show that, unlike NFV-based designs, the limited scalability of current SBA-based 5GC is not inherent to the disaggregated SBA; instead, it stems from (a) the limitations of current implementations (which lack parallelism in key steps of the pipelined execution of microservices) and (b) our oversight regarding the distinctive role these cores play within the network—managing bidirectional (uplink and downlink) events.
Based on these observations, we develop a fast, flexible, and scalable cloud-native 5G core, called Tegra, which proposes scalable and resilient microservices abstractions and state-management strategies to enable autoscaling, load balancing, and fault-tolerance in current SBA-based 5GC. Our evaluation shows that Tegra achieves significantly lower latencies—finishing requests 14× and 4× faster (on average) than Free5GC and Aether, respectively.
An OnRamp blueprint that deploys the Tegra upgrades to SD-Core is planned for the near future.
Team: Bilal Saleem, Omar Basit, Jiayi Meng, Jingqi Huang, Ajay Lotan Thakur, Iftekhar Alam, Christian Maciocco, Y. Charlie Hu, and Muhammad Shahbaz
Telesa: Evolving Mobile Cloud Gaming with 5G Standalone Network Telemetry
Mobile cloud gaming places the simultaneous demands of high capacity and low latency on the wireless network, demands that Private and Metropolitan-Area Standalone 5G networks are poised to meet. However, lacking introspection into the 5G Radio Access Network (RAN), cloud gaming servers are ill-poised to cope with the vagaries of the wireless last hop to a mobile client, while 5G network operators run mostly closed networks, limiting their potential for co-design with the wider internet and user applications. This paper presents Telesa, a passive, incrementally-deployable, and independently-deployable Standalone 5G network telemetry system that streams fine-grained RAN capacity, latency, and retransmission information to application servers to enable better millisecond scale, application-level decisions on offered load and bit rate adaptation than end-to-end latency measurements or end-to-end packet losses currently permit. We design, implement, and evaluate a Telesa telemetry-enhanced game streaming platform, demonstrating exact congestion-control that can better adapt game video bitrate while simultaneously controlling end-to-end latency, thus maximizing game quality of experience.
Our experimental evaluation on the Aether Onramp 5G Standalone software defined core network with the Mosolabs Private 5G small cell demonstrates a 178-249% Quality of Experience improvement versus two state-of-the-art cloud gaming applications.
Team: Haoran Wan, Kyle Jamieson (Princeton Advanced Wireless Systems lab)
Preprint (PDF): arXiv:2402.04454v1
Practical Load Balancing Algorithm for 5G Small Cell Networks Based on Real-World 5G Traffic and O-RAN Architecture
In the 5G network, small cells play a key role in enhancing network capacity and providing ultra-fast, low-latency data services. The traffic load per small cell is highly dynamic and suffers from rapid fluctuations, necessitating appropriate user association algorithms for load balancing. Nevertheless, existing researches often adopt overly tractable traffic models or rely on outdated datasets, making their sophisticated algorithms ineffective. In this paper, we take an improved approach: (i) we collect real-world 5G traffic data ourselves to accurately model the traffic load, and (ii) propose a low-complexity, one-shot user association algorithm, which makes a decision with a single computation without iterative operations. The proposed algorithm determines user association for cell-edge users based on the biased received power of small cells. Crucially, it dynamically adjusts the bias of each small cell in response to its traffic load. We have discovered that the 5G traffic load is dynamic, yet sufficiently predictable due to its daily pattern. Leveraging this characteristic, the proposed algorithm increases the bias value for small cells predicted to have lower traffic load in the near future and decreases it for those expected to have higher traffic load, thus achieving load balancing in 5G networks. The simulation results verify the superiority of our approach in terms of load fairness over the conventional static bias schemes, such as cell range expansion (CRE), although it slightly underperforms compared to the near-optimal results derived from an iterative solution with significantly higher computational complexity.
Team: Hyunmin Yoo, Sangyeon Lee, Geon Kim, Sungjin Lee, Hyuksun Kwon, Hoseong Choi (Kyunghee University Mobile Communication Lab)
Preprint (PDF): https://ieeexplore.ieee.org/document/10659897
User Association and Load Balancing Based on Monte Carlo Tree Search
The user association algorithm for 5G ultra-dense heterogeneous networks (UD-HetNets) comprising multi-tier base stations is becoming increasingly complex. In UD-HetNets, small base stations (SBSs) play an important role in offloading data traffic of user equipments (UEs) requiring high data rate from macro base stations (MBSs) to enhance the quality of services (QoS) of them. However, the traditional cell range expansion (CRE) scheme poses a risk of congestion in certain SBSs and the emergence of UEs monopolizing resources in less congested SBSs, which causes SBS load imbalance and decreases fairness performance. At the same time, determining the optimal user association result for load balancing, considering all possible combinations of associations between UEs and SBSs, leads to prohibitively high computational complexity. To obtain a near-optimal user association solution with manageable computational complexity, in this paper, we propose a heuristic algorithm based on Monte Carlo tree search (MCTS) for user association in UD-HetNet. We model the user association problem as a combinatorial optimization problem and provide a detailed design of the MCTS steps to solve this NP-hard problem. The MCTS algorithm obtains a near-optimal UEs-SBSs combination in terms of load balancing and maximizes the fairness of the overall network. This combination derived from the proposed algorithm aims to achieve load balancing among SBSs and mitigate resource monopolization among UEs. The simulation results show that the proposed algorithm outperforms conventional user association schemes in terms of fairness. As a result, compared to traditional CRE schemes, the proposed method can provide good performance to the UEs receiving data rates of the bottom 50%. Furthermore, the gap between optimal and heuristic solutions does not exceed 4%. Due to its manageable computational complexity, the proposed algorithm can be implemented as an xApp on the O-RAN near-real-time RAN intelligent controller (RIC).
Team: Hyunmin Yoo, Sangyeon Lee, Geon Kim, Sungjin Lee, Hyuksun Kwon, Hoseong Choi (Kyunghee University Mobile Communication Lab)
Preprint (PDF): https://ieeexplore.ieee.org/document/10310137
Load balancing algorithm running on Open RAN RIC
The O-RAN alliance has received attention by presenting an O-RAN architecture. They standardizes wireless interfaces to solve the compatibility problem between multi- vendor of existing radio access network (RAN) architecture. They also disclose an open source framework, which is applicable to programmable base station equipment. In this paper, we analyze the xApp and the near-real time RAN intelligent controller (RIC) serviced by the O-RAN-compatible SD-RAN platform, developed by the open networking foundation (ONF), and address simple simulation results.
Team: Hyunmin Yoo, Sangyeon Lee, Geon Kim, Sungjin Lee, Hyuksun Kwon, Hoseong Choi (Kyunghee University Mobile Communication Lab)
Preprint (PDF): https://ieeexplore.ieee.org/document/9952635