Presents a distribution-aware scheduling framework for LLM inference that reduces P99 TTLT by 35-50% and TTFT by 34-47% versus SRPT with perfect length knowledge using statistical signals instead of predictions.
hub
Distserve: Disaggregating prefill and decoding for goodput-optimized large language model serving
27 Pith papers cite this work, alongside 15 external citations. Polarity classification is still indexing.
27
Pith papers citing it
15
external citations · Pith
abstract
DistServe improves the performance of large language models (LLMs) serving by disaggregating the prefill and decoding computation. Existing LLM serving systems colocate the two phases and batch the computation of prefill and decoding across all users and requests. We find that this strategy not only leads to strong prefill-decoding interferences but also couples the resource allocation and parallelism plans for both phases. LLM applications often emphasize individual latency for each phase: time to first token (TTFT) for the prefill phase and time per output token (TPOT) of each request for the decoding phase. In the presence of stringent latency requirements, existing systems have to prioritize one latency over the other, or over-provision compute resources to meet both. DistServe assigns prefill and decoding computation to different GPUs, hence eliminating prefill-decoding interferences. Given the application's TTFT and TPOT requirements, DistServe co-optimizes the resource allocation and parallelism strategy tailored for each phase. DistServe also places the two phases according to the serving cluster's bandwidth to minimize the communication caused by disaggregation. As a result, DistServe significantly improves LLM serving performance in terms of the maximum rate that can be served within both TTFT and TPOT constraints on each GPU. Our evaluations show that on various popular LLMs, applications, and latency requirements, DistServe can serve 7.4x more requests or 12.6x tighter SLO, compared to state-of-the-art systems, while staying within latency constraints for > 90% of requests.
hub tools
citation-role summary
background 3
citation-polarity summary
roles
background 3polarities
background 3representative citing papers
Disaggregated inference is modeled as three games whose price of anarchy rises at GPU saturation; an adaptive controller reduces the empirical PoA-hat by up to 3.1x on real clusters at modest throughput cost.
ViBE co-optimizes expert placement with measured GPU performance variability in MoE inference to cut execution-time imbalance, delivering 14% better SLO attainment and up to 45% lower P90 TTFT.
Power capping is illusory in LLM decode as memory-bound operation leaves power headroom untouched on 700 W GPUs, while SM clock locking saves up to 32% energy and three DVFS classes appear across attention types.
Hosted open-weight LLM APIs function as time-varying heterogeneous services rather than fixed model artifacts, with demand concentrated, supply-use mismatches, and task-specific routing yielding major cost and throughput gains.
KV-cache eviction, prompt compression, recurrent state bounding, and agent memory consolidation are unified as one rate-distortion problem with a shared lower bound, shared failure mode, and transferable mechanisms.
LLM serving should triage by five-resource analytical floors and wall ordering, not grid search; on 16×H20, TP16 is capacity-capped at ~70 while EP+DP attention reaches ~644 concurrent 8K requests.
OmniPilot combines conformal quantile regression with OOD detection to rank LLM serving configurations on mixed GPUs, reporting 6.2% MAPE throughput prediction and 95% top-1 accuracy on 460 benchmark runs while abstaining on unsupported cases.
KernelSight-LM simulates LLM inference at kernel granularity with cross-generation (12.1% per-kernel error) and target-measured (3.8% error) tiers, yielding end-to-end median errors of 15.4%/12.8%/3.0% and 14.3%/6.2%/2.7% for TTFT/TPOT/throughput across six model families.
Vortex provides a programmable frontend and backend for sparse attention in LLM serving, delivering up to 3.46x throughput over full attention while preserving accuracy.
EnergyLens predicts multi-GPU LLM inference energy consumption with 9-13% MAPE and identifies configurations with up to 52x energy efficiency differences.
Chakra introduces a standardized graph-based execution trace representation for distributed ML workloads along with supporting tools to enable benchmarking, analysis, generation, and co-design across simulators and hardware.
Dooly reduces LLM inference profiling GPU-hours by 56.4% across 12 models while keeping simulation MAPE under 5% for TTFT and 8% for TPOT by making profiling configuration-agnostic and redundancy-aware.
BalanceRoute uses a piecewise-linear F-score (with optional short lookahead) for sticky request routing in LLM serving, reducing DP imbalance and raising end-to-end throughput versus vLLM baselines on production and Azure traces.
A queueing model derives stability conditions for LLM inference services under combined compute and KV cache memory limits, with experimental validation showing typical deviations under 10%.
MemExplorer optimizes heterogeneous memory systems for agentic LLM inference on NPUs and reports up to 2.3x higher energy efficiency than baselines under fixed power budgets.
HybridFlow combines single- and multi-controller paradigms with a 3D-HybridEngine to deliver 1.53x to 20.57x higher throughput for various RLHF algorithms compared to prior systems.
Omni-Flow introduces a three-layer abstraction (Control Flow, Data Flow, Compute Flow) for unified orchestration and KV cache sharing in multimodal inference pipelines.
Organizes the heterogeneous LLM prefill-decode design space along four axes and extracts three boundary decisions with guidance on precision, KV representation, and ownership.
AlignedServe uses prefix-aware batching, large CPU in-flight request pools, batch scheduling, and GPU-to-GPU KV prefetching to raise decoding throughput up to 1.98x and cut latency up to 7.4x versus prior serving systems.
ODMA raises KV-cache utilization by up to 19.25% and throughput by 23-27% on Cambricon MLU accelerators by dynamically adjusting prediction buckets and using a safety pool for LLM serving.
ServeGen characterizes production LLM inference workloads across model types and generates realistic per-client composed workloads that reduce under-provisioning by 50% in a production validation.
ITME uses CXL-hybrid memories for byte-addressable remote memory expansion in LLM inference, achieving up to 35.7% throughput improvement over conventional CPU-offloading.
The paper defines Computational Token Economics and introduces the Token Economics Trilemma as a framework for trade-offs in granularity, real-time performance, and optimality, while outlining a research agenda for three challenge areas.
citing papers explorer
-
Beyond Prediction: Tail-Aware Scheduling for LLM Inference
Presents a distribution-aware scheduling framework for LLM inference that reduces P99 TTLT by 35-50% and TTFT by 34-47% versus SRPT with perfect length knowledge using statistical signals instead of predictions.
-
The Price of Anarchy in Disaggregated Inference
Disaggregated inference is modeled as three games whose price of anarchy rises at GPU saturation; an adaptive controller reduces the empirical PoA-hat by up to 3.1x on real clusters at modest throughput cost.
-
ViBE: Co-Optimizing Workload Skew and Hardware Variability for MoE Serving
ViBE co-optimizes expert placement with measured GPU performance variability in MoE inference to cut execution-time imbalance, delivering 14% better SLO attainment and up to 45% lower P90 TTFT.
-
The Illusion of Power Capping in LLM Decode: A Phase-Aware Energy Characterisation Across Attention Architectures
Power capping is illusory in LLM decode as memory-bound operation leaves power headroom untouched on 700 W GPUs, while SM clock locking saves up to 32% energy and three DVFS classes appear across attention types.
-
When Is the Same Model Not the Same Service? A Measurement Study of Hosted Open-Weight LLM APIs
Hosted open-weight LLM APIs function as time-varying heterogeneous services rather than fixed model artifacts, with demand concentrated, supply-use mismatches, and task-specific routing yielding major cost and throughput gains.
-
What to Keep, What to Forget: A Rate--Distortion View of Memory Compaction in LLMs and Agents
KV-cache eviction, prompt compression, recurrent state bounding, and agent memory consolidation are unified as one rate-distortion problem with a shared lower bound, shared failure mode, and transferable mechanisms.
-
Think Before You Grid-Search: Floor-First Triage for LLM Serving
LLM serving should triage by five-resource analytical floors and wall ordering, not grid search; on 16×H20, TP16 is capacity-capped at ~70 while EP+DP attention reaches ~644 concurrent 8K requests.
-
OmniPilot: An Uncertainty-Aware LLM Inference Advisor for Heterogeneous GPU Clusters
OmniPilot combines conformal quantile regression with OOD detection to rank LLM serving configurations on mixed GPUs, reporting 6.2% MAPE throughput prediction and 95% top-1 accuracy on 460 benchmark runs while abstaining on unsupported cases.
-
KernelSight-LM: A Kernel-Level LLM Inference Simulator
KernelSight-LM simulates LLM inference at kernel granularity with cross-generation (12.1% per-kernel error) and target-measured (3.8% error) tiers, yielding end-to-end median errors of 15.4%/12.8%/3.0% and 14.3%/6.2%/2.7% for TTFT/TPOT/throughput across six model families.
-
Vortex: Efficient and Programmable Sparse Attention Serving for AI Agents
Vortex provides a programmable frontend and backend for sparse attention in LLM serving, delivering up to 3.46x throughput over full attention while preserving accuracy.
-
EnergyLens: Predictive Energy-Aware Exploration for Multi-GPU LLM Inference Optimization
EnergyLens predicts multi-GPU LLM inference energy consumption with 9-13% MAPE and identifies configurations with up to 52x energy efficiency differences.
-
MLCommons Chakra: Advancing Performance Benchmarking and Co-design using Standardized Execution Traces
Chakra introduces a standardized graph-based execution trace representation for distributed ML workloads along with supporting tools to enable benchmarking, analysis, generation, and co-design across simulators and hardware.
-
Dooly: Configuration-Agnostic, Redundancy-Aware Profiling for LLM Inference Simulation
Dooly reduces LLM inference profiling GPU-hours by 56.4% across 12 models while keeping simulation MAPE under 5% for TTFT and 8% for TPOT by making profiling configuration-agnostic and redundancy-aware.
-
Tackling the Data-Parallel Load Balancing Bottleneck in LLM Serving: Practical Online Routing at Scale
BalanceRoute uses a piecewise-linear F-score (with optional short lookahead) for sticky request routing in LLM serving, reducing DP imbalance and raising end-to-end throughput versus vLLM baselines on production and Azure traces.
-
A Queueing-Theoretic Framework for Stability Analysis of LLM Inference with KV Cache Memory Constraints
A queueing model derives stability conditions for LLM inference services under combined compute and KV cache memory limits, with experimental validation showing typical deviations under 10%.
-
MemExplorer: Navigating the Heterogeneous Memory Design Space for Agentic Inference NPUs
MemExplorer optimizes heterogeneous memory systems for agentic LLM inference on NPUs and reports up to 2.3x higher energy efficiency than baselines under fixed power budgets.
-
HybridFlow: A Flexible and Efficient RLHF Framework
HybridFlow combines single- and multi-controller paradigms with a 3D-HybridEngine to deliver 1.53x to 20.57x higher throughput for various RLHF algorithms compared to prior systems.
-
Omni-Flow: A Unified Workflow Orchestration and Distributed KV Cache Sharing Framework for Multimodal Inference
Omni-Flow introduces a three-layer abstraction (Control Flow, Data Flow, Compute Flow) for unified orchestration and KV cache sharing in multimodal inference pipelines.
-
Demystifying the Design Space and Best Practices for Heterogeneous LLM Inference and Serving
Organizes the heterogeneous LLM prefill-decode design space along four axes and extracts three boundary decisions with guidance on precision, KV representation, and ownership.
-
AlignedServe: Orchestrating Prefix-aware Batching to Build a High-throughput and Computing-efficient LLM Serving System
AlignedServe uses prefix-aware batching, large CPU in-flight request pools, batch scheduling, and GPU-to-GPU KV prefetching to raise decoding throughput up to 1.98x and cut latency up to 7.4x versus prior serving systems.
-
ODMA: On-Demand Memory Allocation Strategy for LLM Serving on LPDDR-Class Accelerators
ODMA raises KV-cache utilization by up to 19.25% and throughput by 23-27% on Cambricon MLU accelerators by dynamically adjusting prediction buckets and using a safety pool for LLM serving.
-
ServeGen: Workload Characterization and Generation of Large Language Model Serving in Production
ServeGen characterizes production LLM inference workloads across model types and generates realistic per-client composed workloads that reduce under-provisioning by 50% in a production validation.
-
ITME: Inference Tiered Memory Expansion with Disaggregated CXL-Hybrid Memories
ITME uses CXL-hybrid memories for byte-addressable remote memory expansion in LLM inference, achieving up to 35.7% throughput improvement over conventional CPU-offloading.
-
Computational Challenges in Token Economics: Bridging Economic Theory and AI System Design
The paper defines Computational Token Economics and introduces the Token Economics Trilemma as a framework for trade-offs in granularity, real-time performance, and optimality, while outlining a research agenda for three challenge areas.
-
Token-Operations-Oriented Inference Optimization Techniques for Large Models
A survey of large-model inference optimization, organized as a four-layer 'token-operations' taxonomy: multi-model fusion, model optimization, compute-model fusion, and compute-network-model fusion.
-
A Survey on Efficient Inference for Large Language Models
The paper surveys techniques to speed up and reduce the resource needs of LLM inference, organized by data-level, model-level, and system-level changes, with comparative experiments on representative methods.
- Seeing is Coding: On the Effectiveness of Vision Language Models in Code Understanding