The network verification system Hoyan has been deployed for Alibaba Cloud's wide-area network (WAN) for years and achieved considerable success in preventing misconfiguration-caused network incidents. However, recent years have seen the emergence of new challenges in scalability, usability, and accuracy for Hoyan. This paper presents the new evolution of Hoyan to address these challenges. First, to support the large increase in the number of routers and prefixes on our WAN, Hoyan's simulation has evolved from a centralized fashion to a distributed framework, which improves the efficiency by 5 times and can scale to O(104) routers, millions of prefixes, and billions of flows. Second, to improve Hoyan's usability in checking route change intents, we developed a specification language RCL, which supports the easy specification and automatic verification of route change intents. Third, to ensure high accuracy we enhanced Hoyan's accuracy diagnosis framework, which helped us identify and fix dozens of implementation and modeling issues. Hoyan is used on a daily basis for our WAN. It supports O(100) verification requests each week, prevents O(10) incidents each year, and helps reduce the percentage of misconfiguration-caused network incidents from 56% to 5%.

With the development of intelligent technology, data exhibits characteristics of multi-label and imbalanced distribution, which lead to the degradation of classification model performance. Therefore, addressing multi-label class imbalance has become a hot research topic. Nowadays, over-sampling approaches aim to generate a superset of the original dataset to deal with imbalanced data. However, traditional over-sampling methods only employ the central data point and its nearest neighbor samples to synthesize samples without considering the impact of data distribution. To address these issues, in this paper, we propose an ensemble multi-label over-sampling algorithm (MLCIO) based on Chebyshev inequality and a group optimization strategy. Firstly, to generate more representative and diverse samples, with the seed sample serving as the sphere’s center, Chebyshev inequality is utilized to ensure that synthetic samples fall within its m times the standard deviation. Secondly, a group optimization ranking weighting approach is employed to obtain more reliable and stable label information. Finally, comparative experiments are conducted on 11 imbalanced datasets from various domains using different evaluation metrics. The results demonstrate that our proposal achieves better performance than other approaches.

In this article, we consider the problem of joint transceiver design for millimeter-wave (mmWave)/terahertz (THz) multiuser MIMO integrated sensing and communication (ISAC) systems. Such a problem is formulated into a nonconvex optimization problem, with the objective of maximizing a weighted sum of communication users’ rates and the passive radar’s signal-to-clutter-and-noise ratio (SCNR). By exploring a low-dimensional subspace property of the optimal precoder, a low-dimensional subspace property-inspired block-coordinate-descent (LS-BCD)-based algorithm is proposed with remarkably reduced computational complexity. Our analysis reveals that the hybrid analog/digital beamforming structure can attain the same performance as that of a fully digital precoder, provided that the number of radio frequency (RF) chains is no less than the number of resolvable signal paths. Also, through expressing the precoder as a sum of a communication-precoder and a sensing-precoder, we develop an analytical solution to the joint transceiver design problem by generalizing the idea of block diagonalization (BD) to the ISAC system. Simulation results show that with a proper tradeoff parameter, the proposed methods can achieve a decent compromise between communication and sensing, where the performance of each communication/sensing task experiences only a mild performance loss as compared with the performance attained by optimizing exclusively for a single task.

This paper presents YU, the first verification system for checking traffic load properties under arbitrary failure scenarios that can scale to production Wide Area Networks (WANs). Building a practical YU requires us to address two challenges in terms of generality and efficiency. The state-of-the-art efforts either assume shortest-path-based forwarding (e.g., QARC) or only target single-failure reasoning (e.g., Jingubang). As a result, the former inherently cannot generalize to widely used protocols (e.g., SR and iBGP) that are beyond shortest-path forwarding, while the latter cannot efficiently handle arbitrary failure scenarios. For the generality challenge, we propose an approach inspired by symbolic execution, called symbolic traffic execution, to model the forwarding behavior of a range of practically deployed protocols (e.g., eBGP, iBGP, iGP, and SR) under failure scenarios. For the efficiency challenge, we propose diverse equivalence classification techniques (i.e., k-failure-equivalence and link-local-equivalence reduction) to reduce the symbolic traffic execution overhead caused by both the large size of the production WAN and the huge number of traffic flows traversing it. YU has been used in the daily verification of our WAN for several months and has successfully identified potential failure scenarios that would lead to traffic load violations.

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.