Bodun Hu, Shuozhe Li, Saurabh Agarwal, Myungjin Lee, Akshay Jajoo, Jiamin Li, Le Xu, Geon-Woo Kim, Donghyun Kim, Hong Xu, Amy Zhang, Aditya Akella. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

By aggregating multiple channels, Carrier Aggregation (CA) is an important technology for boosting cellular network bandwidth. Given diverse radio bands made available in 5G networks, CA plays a particularly critical role in achieving the goal of multi-Gbps throughput performance. In this paper, we carry out a timely comprehensive measurement study of CA deployment in commercial 5G networks (as well as 4G networks). We identify the key factors that influence whether CA is deployed and when, as well as which band combinations are used. Thus, we reveal the challenges posed by CA in 5G performance analysis and prediction as well as their implications in application quality-of-experience (QoE). We argue for and develop a novel CA-aware deep learning framework, dubbed Prism5G, which explicitly accounts for the complexity introduced by CA to more effectively predict 5G network throughput performance. Through extensive evaluations, we demonstrate the superiority of Prism5G over existing throughput prediction algorithms. Prism5G improves 5G throughput prediction accuracy by over 14% on average and a maximum of 22%. Using two use cases as examples, we further illustrate how Prism5G can aid applications in optimizing QoE performance.

The ability to accurately estimate job runtime properties allows a scheduler to effectively schedule jobs. State-of-the-art online cluster job schedulers use history-based learning, which uses past job execution information to estimate the runtime properties of newly arrived jobs. However, with fast-paced development in cluster technology (in both hardware and software) and changing user inputs, job runtime properties can change over time, which lead to inaccurate predictions. In this paper, we explore the potential and limitation of real-time learning of job runtime properties, by proactively sampling and scheduling a small fraction of the tasks of each job. Such a task-sampling-based approach exploits the similarity among runtime properties of the tasks of the same job and is inherently immune to changing job behavior. Our analytical and experimental analysis of 3 production traces with different skew and job distribution shows that learning in space can be substantially more accurate. Our simulation and testbed evaluation on Azure of the two learning approaches anchored in a generic job scheduler using 3 production cluster job traces shows that despite its online overhead, learning in space reduces the average Job Completion Time (JCT) by 1.28×, 1.56×, and 1.32× compared to the prior-art history-based predictor.

Outdated and inaccurate channel state information (CSI) prevents the transmitter from adapting to current channel conditions and impairs reliable communication. This paper proposes a deep learning-based methodology for real-time prediction of future CSI, which enables it to be fed back in advance and mitigates the effect of the feedback and processing delay. Predicting the future CSI at the user equipment may also help the gNodeB (gNB) send reference signals less frequently and may help reduce the reference signal transmission overhead. Our approach utilizes Long Short-Term Memory (LSTM) type recurrent neural networks (RNN) to predict the CSI from past channel estimations. However, the knowledge of past channel gains is assumed to be limited, and only partial channel data is available. We propose a new algorithm that utilizes LSTMs and remedies incomplete channel history problem, which we denote by RNN-I. We evaluate the performance of the proposed framework by comparing it with two conventional techniques: persistence and auto-regressive (AR) model-based prediction methods. The results show that the proposed RNN-I algorithm performs well under strong channel mismatch scenarios, while AR methods diverge and can not keep up with fast varying channels if the mismatch in the channel characteristics used in training and inference is significant.

Coflow scheduling improves data-intensive application performance by improving their networking performance. State-of-the-art online coflow schedulers in essence approximate the classic Shortest-Job-First (SJF) scheduling by learning the coflowsizeonline. In particular, they use multiple priority queues to simultaneously accomplish two goals: to sieve long coflows from short coflows, and to schedule short coflows with high priorities. Such a mechanism pays high overhead in learning the coflow size: moving a large coflow across the queues delays small and other large coflows, and moving similar-sized coflows across the queues results in inadvertent round-robin scheduling. We propose Philae, a new online coflow scheduler that exploits the spatial dimension of coflows,i.e.,a coflow has many flows, to drastically reduce the overhead of coflow sizelearning. Philae pre-schedules sampled flows of each coflow and uses their sizes to estimate the average flow size of the coflow. It then resorts to Shortest Coflow First, where the notion of shortest is determined using the learned coflow sizes and coflow contention. We show that the sampling-based learning is robust to flow size skew and has the added benefit of much improved scalability from reduced coordinator-local agent interactions. Our evaluation using an Azure testbed, a publicly available production cluster trace from Facebook shows that compared to the prior art Aalo, Philae reduces the coflow completion time (CCT) in average (P90) cases by$1.50\times $($8.00\times $) on a 150-node testbed and$2.72\times $($9.78\times $) on a 900-node testbed. Evaluation using additional traces further demonstrates Philae’s robustness to flow size skew.

CoFlow scheduling improves data-intensive application performance by improving their networking performance. State-of-the-art CoFlow schedulers in essence approximate the classic online Shortest-Job-First (SJF) scheduling, designed for a single CPU, in a distributed setting, with no coordination among how the flows of a CoFlow at individual ports are scheduled, and as a result suffer two performance drawbacks: (1) The flows of a CoFlow may suffer the out-of-sync problem -- they may be scheduled at different times and become drifting apart, negatively affecting the CoFlow completion time (CCT); (2) FIFO scheduling of flows at each port bears no notion of SJF, leading to suboptimal CCT.