arXiv subjects

Mingxing Zhang

Publications and source records attributed to Mingxing Zhang.

At least 19 recordsLinked to original sources

arXiv2026

AFD-Ledger: Deployment Provisioning for Attention--FFN Disaggregation

Attention--Feed-Forward Network (FFN) Disaggregation (AFD) is emerging as a promising architecture for serving Mixture-of-Experts (MoE) language models. While existing AFD systems improve the efficiency of disaggregated execution, they leave a deployment question unanswered: under the same model, workload, time-per-output-token (TPOT) service-level objective (SLO), hardware budget, hardware catalog, and runtime capabilities, does AFD provide higher throughput than the best collocated deployment? Answering this question requires jointly optimizing hardware assignment and deployment organization for both architectures, making exhaustive provisioning prohibitively expensive. We present AFD-Ledger, an offline analytical provisioning system that independently provisions AFD and collocated deployments using an analytical execution model and an evaluation-bounded hardware search. Across deployment spaces where exhaustive provisioning is feasible, AFD-Ledger reduces complete deployment evaluations by 68.8%--83.5% while still recovering the globally optimal deployment. On three physical LongCat 2.0 deployments, it preserves the correct architecture decision while predicting AFD-to-collocated throughput within 6.6%--9.6% of measurement. Using this validated framework, we show that homogeneous AFD improves fixed-budget throughput in only a minority of the studied settings, heterogeneous AFD requires deployment-level hardware complementarity rather than heuristic device selection, and role-specific hardware improvements matter primarily when they enable better deployment organizations by crossing deployment capability--price boundaries.

cs.DC↗

arXiv2026

TIDE-MC: Two-Sided Interpolative Decomposition for Billion-Scale GPU Matrix Completion

Matrix completion supports large-scale recommendation and scientific computing, yet existing GPU solvers commonly assume that the observed matrix or its dense factors fit in device memory. On real workloads, this assumption leads to out-of-memory failures or severe PCIe overhead under naive paging. We present TIDE-MC, a bounded-memory GPU framework built on Two-Sided Interpolative Decomposition (TSID). TSID uses a sampled template submatrix as an anchor for reconstructing the full low-rank matrix, allowing computation and storage to scale with the template and active data chunks rather than the complete matrix. TIDE-MC realizes this formulation through two execution stages. First, a conflict-free synchronization engine recovers the template using parallel factorization and hierarchical gradient aggregation. Second, a chunked reconstruction pipeline extends the recovered template to the remaining matrix while overlapping PCIe transfers with GPU computation. An asymmetric gradient-clipping scheme stabilizes mixed-precision Tensor Core execution. Across 15 benchmarks, TIDE-MC completes workloads that cause existing GPU solvers to run out of memory. Compared with the evaluated state-of-the-art baselines, it achieves up to 11,647x speedup, reduces peak memory usage by up to 8.5x, and lowers reconstruction error by up to 99.7%. These results show that template-anchored decomposition and stage-specific GPU execution can scale matrix completion beyond device-memory capacity.

cs.DC↗

arXiv2026

Surviving Partial Rank Failures in Wide Expert-Parallel MoE Inference

Mixture-of-Experts (MoE) serving relies on wide expert parallelism (EP) to aggregate the memory capacity and bandwidth of many GPUs within one inference instance. This efficiency comes with a systems cost: every decoding step depends on token dispatch and combination across all active EP ranks, so even one rank failure can disrupt the entire service. Existing EP stacks handle such failures poorly because they treat membership as a fixed configuration established at initialization. The same rank set determines communicator state, expert placement, and the routing metadata baked into CUDA execution graphs, leaving the system with no way to shrink around a failure while keeping the instance valid. This paper argues that partial-failure tolerance should instead be formulated as a live EP validity problem. We present EEP, a communication and runtime substrate that represents membership as explicit, mutable runtime state. EEP repairs the specific state invalidated by a fault: it restores peer reachability without rebuilding the communication substrate, repairs lost expert coverage through a bandwidth-aware hierarchy, and reintegrates repaired ranks without forcing healthy ranks to recapture their CUDA graphs. We implement EEP in an EP serving stack integrated with SGLang and evaluate it under steady-state serving, failure recovery, and rank reintegration. The results show that explicit mutable membership preserves the steady-state fast path, staying within 4.4% of a fixed-membership DeepEP baseline under static serving, while turning a local rank fault from whole-instance downtime into two bounded interruptions. On a single-rank failure workload, EEP incurs an 11s recovery pause and an 8s reintegration pause, and restores throughput to within 95% of the pre-fault level within 52s, whereas a fixed-membership full-restart baseline remains unavailable until 348s.

cs.DC↗

arXiv2026

Prefill-as-a-Service: KVCache of Next-Generation Models Could Go Cross-Datacenter

Prefill-decode (PD) disaggregation has become the standard architecture for large-scale LLM serving, but in practice its deployment boundary is still determined by KVCache transfer. In conventional dense-attention models, prefill generates huge KVCache traffics that keep prefill and decode tightly coupled within a single high-bandwidth network domain, limiting heterogeneous deployment and resource elasticity. Recent hybrid-attention architectures substantially reduce KVCache size, making cross-cluster KVCache transport increasingly plausible. However, smaller KVCache alone does not make heterogeneous cross-datacenter PD serving practical: real workloads remain bursty, request lengths are highly skewed, prefix caches are unevenly distributed, and inter-cluster bandwidth fluctuates. A naive design that fully externalizes prefill can therefore still suffer from congestion, unstable queueing, and poor utilization. We present Prefill-as-a-Service (PrfaaS), a cross-datacenter serving architecture that selectively offloads long-context prefill to standalone, compute-dense prefill clusters and transfers the resulting KVCache over commodity Ethernet to local PD clusters for decode. Rather than treating reduced KVCache as sufficient, PrfaaS combines model-side KV efficiency with system-side selective offloading, bandwidth-aware scheduling, and cache-aware request placement. This design removes the requirement that heterogeneous accelerators share the same low-latency RDMA fabric, enabling independent scaling of prefill and decode capacity across loosely coupled clusters. In a case study using an internal 1T-parameter hybrid model, a PrfaaS-augmented heterogeneous deployment achieves 54% higher serving throughput and 64% lower P90 TTFT than a homogeneous PD baseline, with approximately 15% throughput gain at equal cost, while consuming only modest cross-datacenter bandwidth.

cs.DC↗

arXiv2026

Tessera: Unlocking Heterogeneous GPUs through Kernel-Granularity Disaggregation

Disaggregation maps parts of an AI workload to different types of GPUs, offering a path to utilize modern heterogeneous GPU clusters. However, existing solutions operate at a coarse granularity and are tightly coupled to specific model architectures, leaving much room for performance improvement. This paper presents Tessera, the first kernel disaggregation system to improve performance and cost efficiency on heterogeneous GPUs for large model inference. Our key insight is that kernels within a single application exhibit diverse resource demands, making them the most suitable granularity for aligning computation with hardware capabilities. Tessera integrates offline analysis with online adaptation by extracting precise inter-kernel dependencies from PTX to ensure correctness, overlapping communication with computation through a pipelined execution model, and employing workload-aware scheduling with lightweight runtime adaptation. Extensive evaluations across five heterogeneous GPUs and four model architectures, scaling up to 16 GPUs, show that Tessera improves serving throughput and cost efficiency by up to 2.3x and 1.6x, respectively, compared to existing disaggregation methods, while generalizing to model architectures where prior approaches do not apply. Surprisingly, a heterogeneous GPU pair under Tessera can even exceed the throughput of two homogeneous high-end GPUs at a lower cost.

cs.DC↗

arXiv2026

TENT: A Declarative Slice Spraying Engine for Performant and Resilient Data Movement in Disaggregated LLM Serving

Modern GPU clusters rely on complex, heterogeneous interconnects. As large language model (LLM) serving shifts toward agentic reasoning, KVCache becomes a first-class mobile asset, driving frequent migrations and massive elephant flows that dominate the execution critical path. Operating Mooncake Transfer Engine (TE) on thousands of GPUs exposed a fundamental flaw in existing frameworks: imperative, early-binding path selection. This rigidity results in state-blind striping that ignores congestion and grey failures, leading to bandwidth stranding. This also produces operational fragility where routine faults require manual intervention. We present TENT, a declarative orchestration engine that decouples transfer intent from physical execution. By abstracting interconnects into a unified resource pool, TENT shifts path resolution from initialization to slice-time late binding. Applications simply declare transfer intents, while TENT dynamically "sprays" fine-grained slices across rails based on real-time telemetry and predictive cost modeling. This orchestration eliminates head-of-line (HoL) blocking and enables transparent, sub-50 ms self-healing by rerouting slices around failures or degradations without application-level intervention. TENT serves as the production data plane for LLM inference and reinforcement learning (RL) pipelines at multiple industrial clusters. Our evaluation shows that TE+ outperforms state-of-the-art baselines, including Mooncake TE, NIXL, and UCCL. In LLM inference with SGLang HiCache, TE+ achieves up to 1.36x higher throughput and 26% lower P90 time-to-first-token (TTFT) than Mooncake TE. In RL pipelines, TENT accelerates parameter updates in Moonshot Checkpoint Engine by 20-26%.

cs.DC↗

arXiv2026

SHAPCA: Consistent and Interpretable Explanations for Machine Learning Models on Spectroscopy Data

In recent years, machine learning models have been increasingly applied to spectroscopic datasets for chemical and biomedical analysis. For their successful adoption, particularly in clinical and safety-critical settings, professionals and researchers must be able to understand and trust the reasoning behind model predictions. However, the inherently high dimensionality and strong collinearity of spectroscopy data pose a fundamental challenge to model explainability. These properties not only complicate model training but also undermine the stability and consistency of explanations, leading to fluctuations in feature importance across repeated training runs. Feature extraction techniques have been used to reduce the input dimensionality; these new features hinder the connection between the prediction and the original signal. This study proposes SHAPCA, an explainable machine learning pipeline that combines Principal Component Analysis (for dimensionality reduction) and Shapely Additive exPlanations (for post hoc explanation) to provide explanations in the original input space, which a practitioner can interpret and link back to the biological components. The proposed framework enables analysis from both global and local perspectives, revealing the spectral bands that drive overall model behaviour as well as the instance-specific features that influence individual predictions. Numerical analysis demonstrated the interpretability of the results and greater consistency across different runs.

cs.LG↗

arXiv2026

DualPath: Breaking the Storage Bandwidth Bottleneck in Agentic LLM Inference

The performance of multi-turn, agentic LLM inference is increasingly dominated by KV-Cache storage I/O rather than computation. In prevalent disaggregated architectures, loading the massive KV-Cache from external storage creates a fundamental imbalance: storage NICs on prefill engines become bandwidth-saturated, while those on decoding engines remain idle. This asymmetry severely constrains overall system throughput. We present DualPath, an inference system that breaks this bottleneck by introducing dual-path KV-Cache loading. Beyond the traditional storage-to-prefill path, DualPath enables a novel storage-to-decode path, in which the KV-Cache is loaded into decoding engines and then efficiently transferred to prefill engines via RDMA over the compute network. DualPath combines this optimized data path -- which inherently avoids network congestion and avoids interference with latency-critical model execution communications -- with a global scheduler that dynamically balances load across prefill and decode engines. Our evaluation on three models with production agentic workloads demonstrates that DualPath improves offline inference throughput by up to 1.87$\times$ on our in-house inference system. It can also improve online serving throughput by an average factor of 1.96$\times$ without violating SLO.

cs.DC↗

arXiv2026

From Prefix Cache to Fusion RAG Cache: Accelerating LLM Inference in Retrieval-Augmented Generation

Retrieval-Augmented Generation enhances Large Language Models by integrating external knowledge, which reduces hallucinations but increases prompt length. This increase leads to higher computational costs and longer Time to First Token (TTFT). To mitigate this issue, existing solutions aim to reuse the preprocessed KV cache of each retrieved chunk to accelerate RAG. However, the lack of cross-chunk contextual information leads to a significant drop in generation quality, leaving the potential benefits of KV cache reuse largely unfulfilled. The challenge lies in how to reuse the precomputed KV cache of chunks while preserving generation quality. We propose FusionRAG, a novel inference framework that optimizes both the preprocessing and reprocessing stages of RAG. In the offline preprocessing stage, we embed information from other related text chunks into each chunk, while in the online reprocessing stage, we recompute the KV cache for tokens that the model focuses on. As a result, we achieve a better trade-off between generation quality and efficiency. According to our experiments, FusionRAG significantly improves generation quality at the same recomputation ratio compared to previous state-of-the-art solutions. By recomputing fewer than 15% of the tokens, FusionRAG achieves up to 70% higher normalized F1 scores than baselines and reduces TTFT by 2.66x-9.39x compared to Full Attention.

cs.CL↗

arXiv2026

ActiShade: Activating Overshadowed Knowledge to Guide Multi-Hop Reasoning in Large Language Models

In multi-hop reasoning, multi-round retrieval-augmented generation (RAG) methods typically rely on LLM-generated content as the retrieval query. However, these approaches are inherently vulnerable to knowledge overshadowing - a phenomenon where critical information is overshadowed during generation. As a result, the LLM-generated content may be incomplete or inaccurate, leading to irrelevant retrieval and causing error accumulation during the iteration process. To address this challenge, we propose ActiShade, which detects and activates overshadowed knowledge to guide large language models (LLMs) in multi-hop reasoning. Specifically, ActiShade iteratively detects the overshadowed keyphrase in the given query, retrieves documents relevant to both the query and the overshadowed keyphrase, and generates a new query based on the retrieved documents to guide the next-round iteration. By supplementing the overshadowed knowledge during the formulation of next-round queries while minimizing the introduction of irrelevant noise, ActiShade reduces the error accumulation caused by knowledge overshadowing. Extensive experiments show that ActiShade outperforms existing methods across multiple datasets and LLMs.

cs.CL↗

arXiv2025

OrchANN: Hierarchical Orchestration for Skewed Out-of-Core Vector Search

At billion scale, approximate nearest neighbor search (ANNS) often becomes an out-of-core problem: the full vector collection and index structures exceed memory capacity, making query performance dominated by SSD accesses and DRAM-SSD data movement. Existing systems struggle to strike a balance between accuracy and efficiency: physical-overlap methods replicate vectors or index entries across partitions, enlarging the SSD-resident index and incurring extra I/O; quantization-based methods reduce memory usage, but their approximate distances are less reliable and often require costly raw-vector reranking from SSD to preserve recall. We present OrchANN (Orchestrated ANN Search), an out-of-core ANNS engine that orchestrates query routing, partition access, and query execution under tight memory constraints. OrchANN stores each cluster as a disjoint SSD partition with scale-aware indexes, while a memory-resident graph abstraction provides logical overlap before SSD access. During serving, OrchANN uses query hotness and cluster priorities from the graph abstraction to prune low-value clusters and improve access locality. Across five datasets under strict memory constraints, OrchANN delivers up to 17.2x higher QPS and 25.0x lower latency than state-of-the-art baselines, while preserving accuracy.

cs.DB↗

arXiv2025

Seer: Online Context Learning for Fast Synchronous LLM Reinforcement Learning

Reinforcement Learning (RL) has emerged as a critical technique for advancing modern Large Language Models (LLMs), yet existing synchronous RL systems face severe performance bottlenecks. The rollout phase, which dominates end-to-end iteration time, suffers from substantial long-tail latency and poor resource utilization due to inherent workload imbalance. We present Seer, a novel context learning RL system that addresses these challenges through a key observation: requests sharing the same prompt exhibit strong similarities in output lengths and response patterns. Leveraging this insight, Seer introduces three coordinated techniques: (1) divided rollout for dynamic load balancing, (2) context-aware scheduling to mitigate long-tail request delays, and (3) adaptive grouped speculative decoding to accelerate generation. These mechanisms work in concert to markedly reduce long-tail latency and improve resource efficiency during rollout. Evaluations on production-grade RL workloads demonstrate that Seer achieves up to 2.04$\times$ end-to-end rollout throughput improvement compared to the state-of-the-art synchronous RL systems, while notably reducing long-tail latency by 72-94%.

cs.DC↗

arXiv2025

TrEnv-X: Transparently Share Serverless Execution Environments Across Different Functions and Nodes

Serverless computing is renowned for its computation elasticity, yet its full potential is often constrained by the requirement for functions to operate within local and dedicated background environments, resulting in limited memory elasticity. To address this limitation, this paper introduces TrEnv-X, a co-designed integration of the serverless platform with the operating system and CXL/RDMA-based remote memory pools. TrEnv-X's core innovations are repurposable sandboxes, which can be shared across different functions to decrease the associated creation overhead, and OS-level memory templates, which enable rapid state restoration from CXL/RDMA-based remote memory pools. To further demonstrate TrEnv-X's versatility, we generalize its design from traditional containers for microVM-based agent workloads and introduce new optimizations, including browser sharing and a page cache bypassing mechanism. Our evaluation shows that TrEnv-X achieves up to 7x reduction in P99 latency and 48% memory savings for container-based functions. When applied to LLM agents, it reduces the P99 latency by up to 58% and memory usage by 61% compared to state-of-the-art systems like E2B.

cs.DC↗

arXiv2025

Mooncake: A KVCache-centric Disaggregated Architecture for LLM Serving

Mooncake is the serving platform for Kimi, a leading LLM service provided by Moonshot AI. It features a KVCache-centric disaggregated architecture that separates the prefill and decoding clusters. It also leverages the underutilized CPU, DRAM, and SSD resources of the GPU cluster to implement a disaggregated cache of KVCache. The core of Mooncake is its KVCache-centric scheduler, which balances maximizing overall effective throughput while meeting latency-related Service Level Objectives (SLOs). Unlike traditional studies that assume all requests will be processed, Mooncake faces challenges due to highly overloaded scenarios. To mitigate these, we developed a prediction-based early rejection policy. Experiments show that Mooncake excels in long-context scenarios. Compared to the baseline method, Mooncake can achieve up to a 525% increase in throughput in certain simulated scenarios while adhering to SLOs. Under real workloads, Mooncake's innovative architecture enables Kimi to handle 75% more requests.

cs.DC↗

arXiv2025

RetroInfer: A Vector Storage Engine for Scalable Long-Context LLM Inference

Recent large language models (LLMs) are rapidly extending their context windows, yet inference throughput lags due to increasing GPU memory and bandwidth demands. This is because the key-value (KV) cache, an intermediate structure storing token representations, grows linearly with context length and requires an iterative linear scan for attention computation. A promising direction to accelerate long-context inference is to exploit attention's inherent sparsity by offloading the KV cache to CPU memory and retrieving only a small subset of tokens important to the current generation step. However, prior sparse attention approaches struggle to balance accuracy and retrieval cost due to varying sparsity patterns and inefficient GPU-CPU memory management. We present RetroInfer, a vector storage engine that realizes a sparsity-based KV cache for long-context inference. RetroInfer introduces an Attention-aWare VEctor index (wave index), which fundamentally improves the tradeoff between attention accuracy and retrieval cost through tripartite attention approximation, accuracy-bound attention estimation, and segmented clustering. We also design the wave buffer, a GPU-CPU buffer manager that assigns computation and manages data across heterogeneous hardware. We evaluate RetroInfer across a range of models and workloads, demonstrating up to 4.4X decoding throughput over full attention at 120K context and up to 12.2X over sparse attention baselines at 1 million tokens -- all while preserving full-attention-level accuracy.

cs.LG↗

arXiv2025

Efficient Graph-Based Approximate Nearest Neighbor Search Achieving: Low Latency Without Throughput Loss

The increase in the dimensionality of neural embedding models has enhanced the accuracy of semantic search capabilities but also amplified the computational demands for Approximate Nearest Neighbor Searches (ANNS). This complexity poses significant challenges in online and interactive services, where query latency is a critical performance metric. Traditional graph-based ANNS methods, while effective for managing large datasets, often experience substantial throughput reductions when scaled for intra-query parallelism to minimize latency. This reduction is largely due to inherent inefficiencies in the conventional fork-join parallelism model. To address this problem, we introduce AverSearch, a novel parallel graph-based ANNS framework that overcomes these limitations through a fully asynchronous architecture. Unlike existing frameworks that struggle with balancing latency and throughput, AverSearch utilizes a dynamic workload balancing mechanism that supports continuous, dependency-free processing. This approach not only minimizes latency by eliminating unnecessary synchronization and redundant vertex processing but also maintains high throughput levels. Our evaluations across various datasets, including both traditional benchmarks and modern large-scale model generated datasets, show that AverSearch consistently outperforms current state-of-the-art systems. It achieves up to 2.1-8.9 times higher throughput at comparable latency levels across different datasets and reduces minimum latency by 1.5 to 1.9 times.

cs.DC↗

arXiv2025

Efficient Heterogeneous Large Language Model Decoding with Model-Attention Disaggregation

Transformer-based large language models (LLMs) exhibit impressive performance in generative tasks but also introduce significant challenges in real-world serving due to inefficient use of the expensive, computation-optimized accelerators. Although disaggregated serving architectures have been proposed to split different phases of LLM inference, the efficiency of decoding phase is still low. This is caused by the varying resource demands of different operators in the transformer-based LLMs. Specifically, the attention operator is memory-intensive, exhibiting a memory access pattern that clashes with the strengths of modern accelerators, especially for long context requests. To enhance the efficiency of LLM decoding, we introduce model-attention disaggregation. This approach leverages a collection of cheap, memory-optimized devices for the attention operator while still utilizing high-end accelerators for other parts of the model. This heterogeneous setup ensures that each component is tailored to its specific workload, maximizing overall performance and cost efficiency. Our comprehensive analysis and experiments confirm the viability of splitting the attention computation over multiple devices. Also, the communication bandwidth required between heterogeneous devices proves to be manageable with prevalent networking technologies. To further validate our theory, we develop and deploy Lamina, an LLM inference system that incorporates model-attention disaggregation in a distributed heterogeneous cluster. Experimental results indicate that Lamina can provide 16.1 ~ 90.1% higher estimated throughput than existing solutions with similar costs.

cs.LG↗

arXiv2025

MoBA: Mixture of Block Attention for Long-Context LLMs

Scaling the effective context length is essential for advancing large language models (LLMs) toward artificial general intelligence (AGI). However, the quadratic increase in computational complexity inherent in traditional attention mechanisms presents a prohibitive overhead. Existing approaches either impose strongly biased structures, such as sink or window attention which are task-specific, or radically modify the attention mechanism into linear approximations, whose performance in complex reasoning tasks remains inadequately explored. In this work, we propose a solution that adheres to the ``less structure'' principle, allowing the model to determine where to attend autonomously, rather than introducing predefined biases. We introduce Mixture of Block Attention (MoBA), an innovative approach that applies the principles of Mixture of Experts (MoE) to the attention mechanism. This novel architecture demonstrates superior performance on long-context tasks while offering a key advantage: the ability to seamlessly transition between full and sparse attention, enhancing efficiency without the risk of compromising performance. MoBA has already been deployed to support Kimi's long-context requests and demonstrates significant advancements in efficient attention computation for LLMs. Our code is available at https://github.com/MoonshotAI/MoBA.

cs.LG↗