The emergence of Large Language Models (LLMs) has necessitated the adoption of distributed training techniques, involving the deployment of thousands of GPUs to train a single model. Unfortunately, the efficiency of large-scale distributed training systems is often suboptimal due to the increased likelihood of hardware errors in high-end GPU products and the heightened risk of network traffic collisions. Specifically, GPUs involved in the same job require periodic synchronization to exchange necessary data, such as gradients, parameters, or activations. As a result, any local hardware failure can disrupt training tasks, and the inability to swiftly identify faulty components leads to a significant waste of GPU resources. Moreover, prolonged communication due to traffic collisions can substantially increase GPU waiting times. To address these challenges, we propose a communication-driven solution, namely the C 4. The key insights of C 4 are twofold. First, the load in distributed training exhibits homogeneous characteristics and is divided into iterations through periodic synchronization, therefore hardware anomalies would incur certain syndrome in collective communication. By leveraging this feature, $\mathbf{C} 4$ can rapidly identify the faulty components, swiftly isolate the anomaly, and restart the task, thereby avoiding resource wastage caused by delays in anomaly detection. Second, the predictable communication model of collective communication, involving a limited number of long-lived flows, allows C 4 to efficiently execute traffic planning, substantially reducing bandwidth competition among these flows. The $\mathbf{C 4}$ has been extensively deployed across real-world production systems in a hyperscale cloud provider, yielding a significant improvement in system efficiency, from 30% to $\mathbf{4 5 \%}$. This enhancement is attributed to a $\mathbf{3 0 \%}$ reduction in error-induced overhead and a 15% reduction in communication costs.

Long-term monitoring of highways in soft soil areas, especially during the postconstruction period, is of great significance to ensure transportation safety and the quality of highway construction. Multitemporal interferometric synthetic aperture radar (MTInSAR) provides an effective tool for soft clay highway monitoring. However, most time-series models used in MTInSAR modeling are empirical mathematical functions, which ignores the physical properties of the observed objects and may limit the accuracy of the retrieved deformation and the understanding of the underground settlement dynamics. We propose a novel InSAR time-series deformation model (namely, NREM) with an emphasis on the rheological mechanisms for soft soil highways and environmental factors (temperature, humidity, and precipitation) to improve the accuracy of the traditional InSAR model and assist in the analyzing the rheological properties of soft soil. The NREM is constructed based on a combination of the seasonal model and the Burgers model introduced from the field of rheology. The primary parameters (i.e., viscosity and elastic modulus) are introduced in the NREM and estimated with the generation of time-series surface deformation. In the real data experiments, two highways are selected as the test areas. The results show that the standard deviations (STDs) of the high-pass deformation, which can reflect the modeling accuracy, derived by the NREM are lower than those of the three traditional models, yielding an improvement of 45% for the Lungui Highway (LH) and 50% for the G1508 Highway (GH). The root mean square errors (RMSEs) for deformation results derived from NREM are estimated to be ±5.1 mm compared with the leveling measurements, which outperforms the traditional models. The obtained rheological parameters can broaden the application of InSAR technology and provide a reference for highway engineering.