arXiv subjects

Shujie Han

Publications and source records attributed to Shujie Han.

5 recordsLinked to original sources

arXiv2026

LUMEN: Coordinated Failure Recovery for Distributed LLM Serving

Modern large language model (LLM) serving clusters distribute inference requests across multiple worker processes on different GPUs, but failures are prevalent at scale. When a worker fails, the cluster simultaneously loses the failed worker's GPU-resident key-value (KV) caches and serving capacity, leaving surviving workers to absorb the redirected traffic while re-running interrupted requests from scratch. Existing fault-tolerant systems either restart interrupted requests from scratch or restore KV caches from checkpoints stored on a fixed neighboring worker, but both approaches route recovery work without considering current cluster load and leave the recovering worker idle during model reload. We present LUMEN, a fault-tolerant LLM serving system that treats recovery as a load-aware coordination problem across three decision points: checkpoint placement before failures, interrupted-request distribution at failure time, and serving capacity restoration during model reload. We evaluate LUMEN using both prototype experiments and large-scale simulations and demonstrate significant improvements in serving and recovery times.

cs.DC↗

arXiv2026

TierCheck: Tiered Checkpointing for Fault Tolerance in Large Language Model Training

Large Language Model (LLM) training is frequently interrupted by a heterogeneous spectrum of failures, from common GPU crashes to catastrophic cluster-wide outages. Existing checkpointing systems rely on monolithic, single-tier storage backend, forcing a trade-off between state-saving overhead and recovery speed. We propose TierCheck, a cluster-aware tiered checkpointing system that aligns storage placement with failure heterogeneity. TierCheck adopts a three-tier design that maintains lightweight differential checkpoints in local and peer memory for fast localized recovery, while asynchronously migrating heavyweight base checkpoints to remote persistent storage. It also ensures strict global consistency across tiers without stalling training, and achieves fast cluster-aware checkpoint restoration during recovery. Evaluations on models up to 40 billion parameters show that TierCheck achieves low training overhead, reduces end-to-end checkpointing time to under 10s, and supports high-frequency checkpointing, ultimately striking an optimal balance between low-overhead persistence and fast recovery.

cs.DC↗

arXiv2025

BSODiag: A Global Diagnosis Framework for Batch Servers Outage in Large-scale Cloud Infrastructure Systems

Cloud infrastructure is the collective term for all physical devices within cloud systems. Failures within the cloud infrastructure system can severely compromise the stability and availability of cloud services. Particularly, batch servers outage, which is the most fatal failure, could result in the complete unavailability of all upstream services. In this work, we focus on the batch servers outage diagnosis problem, aiming to accurately and promptly analyze the root cause of outages to facilitate troubleshooting. However, our empirical study conducted in a real industrial system indicates that it is a challenging task. Firstly, the collected single-modal coarse-grained failure monitoring data (i.e., alert, incident, or change) in the cloud infrastructure system is insufficient for a comprehensive failure profiling. Secondly, due to the intricate dependencies among devices, outages are often the cumulative result of multiple failures, but correlations between failures are difficult to ascertain. To address these problems, we propose BSODiag, an unsupervised and lightweight diagnosis framework for batch servers outage. BSODiag provides a global analytical perspective, thoroughly explores failure information from multi-source monitoring data, models the spatio-temporal correlations among failures, and delivers accurate and interpretable diagnostic results. Experiments conducted on the Alibaba Cloud infrastructure system show that BSODiag achieves 87.5% PR@3 and 46.3% PCR, outperforming baseline methods by 10.2% and 3.7%, respectively.

cs.DC↗

arXiv2024

The Design and Implementation of a High-Performance Log-Structured RAID System for ZNS SSDs

Zoned Namespace (ZNS) defines a new abstraction for host software to flexibly manage storage in flash-based SSDs as append-only zones. It also provides a Zone Append primitive to further boost the write performance of ZNS SSDs by exploiting intra-zone parallelism. However, making Zone Append effective for reliable and scalable storage, in the form of a RAID array of multiple ZNS SSDs, is non-trivial, since Zone Append offloads address management to ZNS SSDs and requires hosts to specifically manage RAID stripes across multiple drives. We propose ZapRAID, a high-performance log-structured RAID system for ZNS SSDs by carefully exploiting Zone Append to achieve high write parallelism and lightweight stripe management. ZapRAID adopts a group-based data layout with a coarse-grained ordering across multiple groups of stripes, such that it can use small-size metadata for stripe management on a per-group basis under Zone Append. It further adopts hybrid data management to simultaneously achieve intra-zone and inter-zone parallelism through a careful combination of both Zone Write and Zone Append primitives. We implement ZapRAID as a user-space block device, and evaluate ZapRAID using microbenchmarks, trace-driven experiments, and real-application experiments. Our evaluation results show that ZapRAID achieves high write throughput and maintains high performance in normal reads, degraded reads, crash recovery, and full-drive recovery.

cs.DC↗

arXiv2019

Robust Data Preprocessing for Machine-Learning-Based Disk Failure Prediction in Cloud Production Environments

To provide proactive fault tolerance for modern cloud data centers, extensive studies have proposed machine learning (ML) approaches to predict imminent disk failures for early remedy and evaluated their approaches directly on public datasets (e.g., Backblaze SMART logs). However, in real-world production environments, the data quality is imperfect (e.g., inaccurate labeling, missing data samples, and complex failure types), thereby degrading the prediction accuracy. We present RODMAN, a robust data preprocessing pipeline that refines data samples before feeding them into ML models. We start with a large-scale trace-driven study of over three million disks from Alibaba Cloud's data centers, and motivate the practical challenges in ML-based disk failure prediction. We then design RODMAN with three data preprocessing echniques, namely failure-type filtering, spline-based data filling, and automated pre-failure backtracking, that are applicable for general ML models. Evaluation on both the Alibaba and Backblaze datasets shows that RODMAN improves the prediction accuracy compared to without data preprocessing under various settings.

cs.LG↗