The rapid growth in both computational density and scale in AI models in recent years motivates the construction of an efficient and reliable dedicated network infrastructure. This paper presents the design, implementation, and operation of Meta's Remote Direct Memory Access over Converged Ethernet (RoCE) networks for distributed AI training. Our design principles involve a deep understanding of the workloads, and we translated these insights into the design of various network components: Network Topology - To support the rapid evolution of generations of AI hardware platforms, we separated GPU-based training into its own "backend" network. Routing - Training workloads inherently impose load imbalance and burstiness, so we deployed several iterations of routing schemes to achieve near-optimal traffic distribution. Transport - We outline how we initially attempted to use DCQCN for congestion management but then pivoted away from DCQCN to instead leverage the collective library itself to manage congestion. Operations - We share our experience operating large-scale AI networks, including toolings we developed and troubleshooting examples.

This paper presents HPN, Alibaba Cloud's data center network for large language model (LLM) training. Due to the differences between LLMs and general cloud computing (e.g., in terms of traffic patterns and fault tolerance), traditional data center networks are not well-suited for LLM training. LLM training produces a small number of periodic, bursty flows (e.g., 400Gbps) on each host. This characteristic of LLM training predisposes Equal-Cost Multi-Path (ECMP) to hash polarization, causing issues such as uneven traffic distribution. HPN introduces a 2-tier, dual-plane architecture capable of interconnecting 15K GPUs within one Pod, typically accommodated by the traditional 3-tier Clos architecture. Such a new architecture design not only avoids hash polarization but also greatly reduces the search space for path selection. Another challenge in LLM training is that its requirement for GPUs to complete iterations in synchronization makes it more sensitive to singlepoint failure (typically occurring on ToR). HPN proposes a new dual-ToR design to replace the single-ToR in traditional data center networks. HPN has been deployed in our production for more than eight months. We share our experience in designing, and building HPN, as well as the operational lessons of HPN in production.

Mobile health represents the future trend of health care due to its great potential in improving health care efficiency, accessibility and quality. It is particularly beneficial for chronic disease patients who require long-term and regular services. To improve the products, mobile health developers need to understand patient needs, values and preferences, and assess the key factors that influence their mobile health adoption intentions. This study focuses on identifying the influential factors of patients’ adoption intention of m-Health. A structural equation model is constructed, and the confirmatory factor analysis and standard path coefficient are used to explore the key factors in chronic disease patients’ adoption process of m-Health. The results show that perceived usefulness and perceived ease of use have the strongest positive effect on patients’ adoption intention. Meanwhile, subjective norm, existing degree of satisfaction, network effect, and cost factor also influence adoption intention. Finally, the House of Quality method is used to examine the relative importance of various properties of m-Health. The Teoriya Resheniya Izobreatatelskikh Zadatch method is applied to resolve the contradictions between these properties. Our study offers important insights for mobile health developers on how to optimally design a product, thereby increasing users’ adoption intention and overall satisfaction.

Electric vehicles (EVs) are now widely acknowledged as a potential ideal means of transportation in the near future in terms of environmental protection and oil crisis. The possible success of the future market for EVs is based on how much of EVs’ value can be perceived by their potential customers. Thus, research on customer perceived value (CPV) of EVs can help us, and especially EV manufacturers, understand the main factors contributing to CPV and how to design suitable EVs that can yield higher CPV. This paper first constructs a multi-scale model for the measurement of CPV based on surveys conducted at Shanghai, China. Then, the decision-making trial and evaluation laboratory method is applied to evaluate the importance of every scale and depict the internal relations among different scales on the impact-relations map (IRM). Further, a novel version of the house of quality is created to conduct technical feasibility analysis for the improvement of each scale. Finally, market segmentation for EV industry is proposed and discussed based on the analysis of the IRM, which could be a practical strategy for EV manufacturers to design appealing EVs and deliver the proper value at the right cost to the right people.

Understanding patient value is a key mechanism in addressing current healthcare challenges, as well as for establishing a patient-centered healthcare system. With high patient dissatisfaction, and the growing tension between patients and providers in China, understanding patient value and evaluating healthcare delivery performance becomes urgent. The objective of this study is to identify patient value dimensions in healthcare and to understand their relationships with patient satisfaction and loyalty. We conduct a case study in two hospitals in Shanghai, China. We identify economic value, health value, fair value, and supplemental value as four value dimensions. We use the combinational weighting method to evaluate the value-added ability of healthcare services, and use the structural equation model to examine the relationships between patient value, satisfaction, and loyalty. The value-added evaluation indicates significant room for improvement related to economic value for the large hospital, as well as to health value and fair value for the community hospital. Our findings inform hospitals and physicians of areas of focus, particularly improving those with low evaluation and high importance, to maximise patient satisfaction and loyalty. We also provide managerial insights for healthcare policy makers, such as strengthening primary health care, ensuring rational pricing, and improving hospital governance.