Imagen de Portada

TD-Pipe: Temporally-Disaggregated Pipeline Parallelism Architecture for High-Throughput LLM Inference

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Zhang, Hongbin, Wei, Taosheng, Zheng, Zhenyi, Du, Jiangsu, Chen, Zhiguang, Lu, Yutong
Formato:	Preprint
Publicado:	2025
Materias:	Distributed, Parallel, and Cluster Computing
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

PipeMax: Enhancing Offline LLM Inference on Commodity GPU Servers
por: Zhang, Hongbin, et al.
Publicado: (2026)

EcoServe: Enabling Cost-effective LLM Serving with Proactive Intra- and Inter-Instance Orchestration
por: Du, Jiangsu, et al.
Publicado: (2025)

gLLM: Global Balanced Pipeline Parallelism System for Distributed LLM Serving with Token Throttling
por: Guo, Tianyu, et al.
Publicado: (2025)

AlignedServe: Orchestrating Prefix-aware Batching to Build a High-throughput and Computing-efficient LLM Serving System
por: Bai, Fengyao, et al.
Publicado: (2026)

Ghidorah: Fast LLM Inference on Edge with Speculative Decoding and Hetero-Core Parallelism
por: Wei, Jinhui, et al.
Publicado: (2025)

SiPipe: Bridging the CPU-GPU Utilization Gap for Efficient Pipeline-Parallel LLM Inference
por: He, Yongchao, et al.
Publicado: (2025)

PipeBoost: Resilient Pipelined Architecture for Fast Serverless LLM Scaling
por: Liu, Chongpeng, et al.
Publicado: (2025)

Boosting LLM Serving through Spatial-Temporal GPU Resource Sharing
por: Lin, Zejia, et al.
Publicado: (2025)

Arrow: Adaptive Scheduling Mechanisms for Disaggregated LLM Inference Architecture
por: Wu, Yu, et al.
Publicado: (2025)

Shift Parallelism: Low-Latency, High-Throughput LLM Inference for Dynamic Workloads
por: Hidayetoglu, Mert, et al.
Publicado: (2025)

DOPD: A Dynamic PD-Disaggregation Architecture for Maximizing Goodput in LLM Inference Serving
por: Liao, Junhan, et al.
Publicado: (2025)

JanusPipe: Efficient Pipeline Parallel Training for Machine Learning Interatomic Potentials
por: Wang, Hongyu, et al.
Publicado: (2026)

Inference without Interference: Disaggregate LLM Inference for Mixed Downstream Workloads
por: Hu, Cunchen, et al.
Publicado: (2024)

PipeLive: Efficient Live In-place Pipeline Parallelism Reconfiguration for Dynamic LLM Serving
por: Bai, Xu, et al.
Publicado: (2026)

From Attention to Disaggregation: Tracing the Evolution of LLM Inference
por: Kumar, Madabattula Rajesh, et al.
Publicado: (2025)

FlexPipe: Adapting Dynamic LLM Serving Through Inflight Pipeline Refactoring in Fragmented Serverless Clusters
por: Lin, Yanying, et al.
Publicado: (2025)

PipeSD: An Efficient Cloud-Edge Collaborative Pipeline Inference Framework with Speculative Decoding
por: Han, Yunhe, et al.
Publicado: (2026)

Efficient Multi-round LLM Inference over Disaggregated Serving
por: He, Wenhao, et al.
Publicado: (2026)

InfiniPipe: Elastic Pipeline Parallelism for Efficient Variable-Length Long-Context LLM Training
por: Wang, Shiju, et al.
Publicado: (2025)

Taming Request Imbalance: SLO-Aware Scheduling for Disaggregated LLM Inference
por: Wang, Qipeng
Publicado: (2026)

KVDirect: Distributed Disaggregated LLM Inference
por: Chen, Shiyang, et al.
Publicado: (2024)

Injecting Adrenaline into LLM Serving: Boosting Resource Utilization and Throughput via Attention Disaggregation
por: Liang, Yunkai, et al.
Publicado: (2025)

OptPipe: Memory- and Scheduling-Optimized Pipeline Parallelism for LLM Training
por: Li, Hongpei, et al.
Publicado: (2025)

CrossPipe: Towards Optimal Pipeline Schedules for Cross-Datacenter Training
por: Chen, Tiancheng, et al.
Publicado: (2025)

CONCUR: High-Throughput Agentic Batch Inference of LLM via Congestion-Based Concurrency Control
por: Chen, Qiaoling, et al.
Publicado: (2026)

HelixPipe: Efficient Distributed Training of Long Sequence Transformers with Attention Parallel Pipeline Parallelism
por: Zhang, Geng, et al.
Publicado: (2025)

Efficient MoE Inference with Fine-Grained Scheduling of Disaggregated Expert Parallelism
por: Pan, Xinglin, et al.
Publicado: (2025)

SPPO:Efficient Long-sequence LLM Training via Adaptive Sequence Pipeline Parallel Offloading
por: Chen, Qiaoling, et al.
Publicado: (2025)

Matrix-PIC: Harnessing Matrix Outer-product for High-Performance Particle-in-Cell Simulations
por: Rao, Yizhuo, et al.
Publicado: (2026)

DiffusionPipe: Training Large Diffusion Models with Efficient Pipelines
por: Tian, Ye, et al.
Publicado: (2024)

Taming the Chaos: Coordinated Autoscaling for Heterogeneous and Disaggregated LLM Inference
por: Li, Rongzhi, et al.
Publicado: (2025)

Disaggregated Prefill and Decoding Inference System for Large Language Model Serving on Multi-Vendor GPUs
por: Chen, Xing, et al.
Publicado: (2025)

Janus: Disaggregating Attention and Experts for Scalable MoE Inference
por: Zhang, Zhexiang, et al.
Publicado: (2025)

Optimizing High-Throughput Distributed Data Pipelines for Reproducible Deep Learning at Scale
por: Mittal, Kashish, et al.
Publicado: (2026)

Balancing Pipeline Parallelism with Vocabulary Parallelism
por: Yeung, Man Tsung, et al.
Publicado: (2024)

PipeInfer: Accelerating LLM Inference using Asynchronous Pipelined Speculation
por: Butler, Branden, et al.
Publicado: (2024)

Lotus: Optimizing Disaggregated Transactions with Disaggregated Locks
por: Hu, Zhisheng, et al.
Publicado: (2025)

SIMPLE: Disaggregating Sampling from GPU Inference into a Decision Plane for Faster Distributed LLM Serving
por: Zhao, Bohan, et al.
Publicado: (2025)

DisagMoE: Computation-Communication overlapped MoE Training via Disaggregated AF-Pipe Parallelism
por: Zeng, Zhichen, et al.
Publicado: (2026)

Amoeba: Runtime Tensor Parallel Transformation for LLM Inference Services
por: Chen, Haoyu, et al.
Publicado: (2025)