Hong Xu

Shuhao Liu, Wei Bai, Hong Xu et al.

Data center networks need to provide low latency, especially at the tail, as demanded by many interactive applications. To improve tail latency, existing approaches require modifications to switch hardware and/or end-host operating systems, making them difficult to be deployed. We present the design, implementation, and evaluation of RepNet, an application layer transport that can be deployed today. RepNet exploits the fact that only a few paths among many are congested at any moment in the network, and applies simple flow replication to mice flows to opportunistically use the less congested path. RepNet has two designs for flow replication: (1) RepSYN, which only replicates SYN packets and uses the first connection that finishes TCP handshaking for data transmission, and (2) RepFlow which replicates the entire mice flow. We implement RepNet on {\tt node.js}, one of the most commonly used platforms for networked interactive applications. {\tt node}'s single threaded event-loop and non-blocking I/O make flow replication highly efficient. Performance evaluation on a real network testbed and in Mininet reveals that RepNet is able to reduce the tail latency of mice flows, as well as application completion times, by more than 50\%.

Yitao Yang, Yangtao Deng, Yifan Xiong et al.

AI workloads incur frequent failures and incidents from the underlying infrastructure. The current incident management workflow follows a provider-centric paradigm, where users report incidents to the infrastructure provider who then conducts troubleshooting. Due to the large number of incidents and the manual nature of the troubleshooting process, the provider often takes several days to resolve an incident, resulting in operational delays and productivity loss. To address these challenges, we present TSGuard, a user-centric multi-agent system that delivers immediate incident diagnosis to users who deploy the workloads. The core innovation of TSGuard is twofold: (1) constructing domain-specific knowledge bases by mining historical on-call experiences in the offline phase, and (2) mimicking human expert diagnosis via structured reasoning and iterative trial-and-error in the online phase. Evaluation using production incident records from Microsoft Azure demonstrates that TSGuard significantly outperforms state-of-the-art baselines, improving diagnostic accuracy by 19.8%. Furthermore, TSGuard reduces the average verification time by 63.4% compared to the sequential execution baseline.

Xin Tan, Yimin Jiang, Yitao Yang et al.

Large language model (LLM)-based applications consist of both LLM and non-LLM components, each contributing to the end-to-end latency. Despite great efforts to optimize LLM inference, end-to-end workflow optimization has been overlooked. Existing frameworks employ coarse-grained orchestration with task modules, which confines optimizations to within each module and yields suboptimal scheduling decisions. We propose fine-grained end-to-end orchestration, which utilizes task primitives as the basic units and represents each query's workflow as a primitive-level dataflow graph. This explicitly exposes a much larger design space, enables optimizations in parallelization and pipelining across primitives of different modules, and enhances scheduling to improve application-level performance. We build Teola, a novel orchestration framework for LLM-based applications that implements this scheme. Comprehensive experiments show that Teola can achieve up to 2.09x speedup over existing systems across various popular LLM applications. The code is available at https://github.com/NetX-lab/Ayo.

Yangtao Deng, Lei Zhang, Qinlong Wang et al.

Reliability is essential for ensuring efficiency in LLM training. However, many real-world reliability issues remain difficult to resolve, resulting in wasted resources and degraded model performance. Unfortunately, today's collective communication libraries operate as black boxes, hiding critical information needed for effective root cause analysis. We propose Mycroft, a lightweight distributed tracing and root cause analysis system designed to address previously hidden reliability issues in collective communication. Mycroft's key idea is to trace collective communication states and leverage internal control and data dependencies to resolve reliability problems in LLM training. Mycroft has been deployed at ByteDance for over six months to debug collective communication related issues at runtime. It detected anomalies within 15 seconds in 90% of cases and identified the root cause within 20 seconds in 60% of cases. We also conducted extensive fault injection experiments to demonstrate Mycroft's capability and efficiency.

Yu Mao, Jingzong Li, Jun Wang et al.

Neural image compression, necessary in various machine-to-machine communication scenarios, suffers from its heavy encode-decode structures and inflexibility in switching between different compression levels. Consequently, it raises significant challenges in applying the neural image compression to edge devices that are developed for powerful servers with high computational and storage capacities. We take a step to solve the challenges by proposing a new transformer-based edge-compute-free image coding framework called Easz. Easz shifts the computational overhead to the server, and hence avoids the heavy encoding and model switching overhead on the edge. Easz utilizes a patch-erase algorithm to selectively remove image contents using a conditional uniform-based sampler. The erased pixels are reconstructed on the receiver side through a transformer-based framework. To further reduce the computational overhead on the receiver, we then introduce a lightweight transformer-based reconstruction structure to reduce the reconstruction load on the receiver side. Extensive evaluations conducted on a real-world testbed demonstrate multiple advantages of Easz over existing compression approaches, in terms of adaptability to different compression levels, computational efficiency, and image reconstruction quality.

Jingyao Li, Pengguang Chen, Xuan Ju et al.

Thanks to advances in deep learning techniques, Human Pose Estimation (HPE) has achieved significant progress in natural scenarios. However, these models perform poorly in artificial scenarios such as painting and sculpture due to the domain gap, constraining the development of virtual reality and augmented reality. With the growth of model size, retraining the whole model on both natural and artificial data is computationally expensive and inefficient. Our research aims to bridge the domain gap between natural and artificial scenarios with efficient tuning strategies. Leveraging the potential of language models, we enhance the adaptability of traditional pose estimation models across diverse scenarios with a novel framework called VLPose. VLPose leverages the synergy between language and vision to extend the generalization and robustness of pose estimation models beyond the traditional domains. Our approach has demonstrated improvements of 2.26% and 3.74% on HumanArt and MSCOCO, respectively, compared to state-of-the-art tuning strategies.

Yu Zhang, Guoming Tang, Qianyi Huang et al.

Non-intrusive load monitoring (NILM) is a well-known single-channel blind source separation problem that aims to decompose the household energy consumption into itemised energy usage of individual appliances. In this way, considerable energy savings could be achieved by enhancing household's awareness of energy usage. Recent investigations have shown that deep neural networks (DNNs) based approaches are promising for the NILM task. Nevertheless, they normally ignore the inherent properties of appliance operations in the network design, potentially leading to implausible results. We are thus motivated to develop the dual Deep Neural Networks (dual-DNN), which aims to i) take advantage of DNNs' learning capability of latent features and ii) empower the DNN architecture with identification ability of universal properties. Specifically in the design of dual-DNN, we adopt one subnetwork to measure power ratings of different appliances' operation states, and the other subnetwork to identify the running states of target appliances. The final result is then obtained by multiplying these two network outputs and meanwhile considering the multi-state property of household appliances. To enforce the sparsity property in appliance's state operating, we employ median filtering and hard gating mechanisms to the subnetwork for state identification. Compared with the state-of-the-art NILM methods, our dual-DNN approach demonstrates a 21.67% performance improvement in average on two public benchmark datasets.

Hongming Huang, Peng Cheng, Hong Xu et al.

We advocate that simulation based on offline profiling is a promising approach to better understand and improve the complex ML systems. Our approach uses operation-level profiling and dataflow based simulation to ensure it offers a unified and automated solution for all frameworks and ML models, and is also accurate by considering the various parallelization strategies in a real system.

Xiaorui Wu, Hong Xu, Honglin Zhang et al.

Production recommendation systems rely on embedding methods to represent various features. An impeding challenge in practice is that the large embedding matrix incurs substantial memory footprint in serving as the number of features grows over time. We propose a similarity-aware embedding matrix compression method called Saec to address this challenge. Saec clusters similar features within a field to reduce the embedding matrix size. Saec also adopts a fast clustering optimization based on feature frequency to drastically improve clustering time. We implement and evaluate Saec on Numerous, the production distributed machine learning system in Tencent, with 10-day worth of feature data from QQ mobile browser. Testbed experiments show that Saec reduces the number of embedding vectors by two orders of magnitude, compresses the embedding size by ~27x, and delivers the same AUC and log loss performance.

Xiaorui Wu, Hong Xu, Bo Li et al.

The parameter server architecture is prevalently used for distributed deep learning. Each worker machine in a parameter server system trains the complete model, which leads to a hefty amount of network data transfer between workers and servers. We empirically observe that the data transfer has a non-negligible impact on training time. To tackle the problem, we design a new distributed training system called Stanza. Stanza exploits the fact that in many models such as convolution neural networks, most data exchange is attributed to the fully connected layers, while most computation is carried out in convolutional layers. Thus, we propose layer separation in distributed training: the majority of the nodes just train the convolutional layers, and the rest train the fully connected layers only. Gradients and parameters of the fully connected layers no longer need to be exchanged across the cluster, thereby substantially reducing the data transfer volume. We implement Stanza on PyTorch and evaluate its performance on Azure and EC2. Results show that Stanza accelerates training significantly over current parameter server systems: on EC2 instances with Tesla V100 GPU and 10Gb bandwidth for example, Stanza is 1.34x--13.9x faster for common deep learning models.