Proactive Edge Drafting

Ivan D. Zyrianoff, Lorenzo Gigli, Federico Montori, L. Sciullo, C. Kamienski, M. D. Felice

The Cloud-to-Things

Ke Luo, Kongyange Zhao, Ouyang Tao, Xiaoxi Zhang, Zhi Zhou, Hao Wang, Xu Chen

Benefiting from hardware upgrades and deep learning techniques, more and more end devices can independently support a variety of intelligent applications. Further powered by edge computing technologies, the end-edge collaboration paradigm becomes one mainstream approach for achieving advanced edge intelligence (EI). To fully exploit the system resources, it is desirable to coordinate diverse EI services efficiently. Thus, we present a novel framework to jointly optimize the cost-performance trade-off for two distinct but typical EI services, where end devices simultaneously perform federated learning (FL) model training and conduct model inference with the assistance of edge offloading. However, balancing the long-term cost-performance trade-off is highly non-trivial, especially in the absence of knowledge of future system dynamics. Moreover, the capacity heterogeneity further increases the difficulty of service coordination among resource-limited end devices. To overcome these challenges, we first analyze the optimality of inference offloading decisions with and without FL model training and quantify their mutual effects due to local resource contention. By incorporating the loss estimation of FL training model, we then propose a novel proactive policy with theoretical guarantees, which proactively controls the stopping of FL training procedure to balance well the trade-offs between FL model performance and resource costs while fulfilling the inference performance requirements. Extensive results show the efficiency and robustness of our proposed algorithm for EI service coordination in dynamic end-edge collaboration scenarios.

Thang Le Duc, Chanh Nguyen, Per-Olov Östberg

Accurate workload prediction is essential for proactive resource allocation in large-scale Content Delivery Networks (CDNs), where traffic patterns are highly dynamic and geographically distributed. This paper introduces a CDN-tailored prediction and autoscaling framework that integrates statistical and deep learning models within an adaptive feedback loop. The framework is evaluated using 18 months of real traffic traces from a production multi-tier CDN, capturing realistic workload seasonality, cache–tier interactions, and propagation delays. Unlike generic cloud-edge predictors, our design incorporates CDN-specific features and model-switching mechanisms to balance prediction accuracy with computational cost. Seasonal ARIMA (S-ARIMA), Long Short-Term Memory (LSTM), Bidirectional LSTM (Bi-LSTM), and Online Sequential Extreme Learning Machine (OS-ELM) are combined to support both short-horizon scaling and longer-term capacity planning. The predictions drive a queue-based resource-estimation model, enabling proactive cache–server scaling with low rejection rates. Experimental results demonstrate that the framework maintains high accuracy while reducing computational overhead through adaptive model selection. The proposed approach offers a practical, production-tested solution for predictive autoscaling in CDNs and can be extended to other latency-sensitive edge-cloud services with hierarchical architectures.

Jinwoo Park, S. Cho, Dongsu Han

Large language models (LLMs) power many modern applications, but serving them at scale remains costly and resource-intensive. Current server-centric systems overlook consumer-grade GPUs at the edge. We introduce SpecEdge, an edge-assisted inference framework that splits LLM workloads between edge and server GPUs using a speculative decoding scheme, exchanging only token outputs over the network. SpecEdge employs proactive edge drafting to overlap edge token creation with server verification and pipeline-aware scheduling that interleaves multiple user requests to increase server-side throughput. Experiments show SpecEdge enhances overall cost efficiency by 1.91x through achieving 2.22x server throughput, and reduces inter token latency by 11.24% compared to a server-only baseline, introducing a scalable, cost-effective paradigm for LLM serving. The code is available at https://github.com/kaist-ina/specedge

Zhoufan Liao, Pang Liu, Bin Zheng, Xiaoyong Tang

Edge caching is a promising technique for effectively reducing backhaul pressure and content access latency in the Internet of Vehicles (IoV). The existing content caching solutions still face the following challenges: 1) contents cached on edge servers are outdated quickly as time and user preferences change; 2) the large amount of vehicle data causes huge communication overheads; and 3) limited storage resources of edge servers. Simultaneously considering these issues to reduce transmission latency is a large-scale 0–1 constraint problem, which is NP-hard, and boosting cache hit rates is a key entry point. In this work, we propose a context-aware proactive caching strategy (CPCS) based on asynchronous federated learning (AFL), which works as follows. To improve the accuracy of content popularity prediction, thus improving the cache hit rate, we combine contextual information between different contents and use long and short-term memory networks to analyze the dynamic preferences of vehicle users. After that, vehicles complete the model training and upload via an asynchronous federation learning to complete the popularity prediction. To explore the problem of local models being outdated in AFL, CPCS integrates model compression algorithms, enhancing system efficiency and prediction accuracy. With the prediction results, CPCS gives a content placement algorithm based on the prediction results to approximate the optimal caching scheme. Simulation results show that the CPCS can improve the cache hit rate by 17% at most compared to existing state-of-the-art caching strategies.