Filtered by: GPU × Workload × Clear all

Eidola: Modeling Multi-GPU Network Communication Traffic in Distributed AI Workloads

Ranganath R. Selagamsetty, Matthew Poremba, Bradford M. Beckmann, Joshua San Miguel 2026-06-14
Gem5 Interconnect Microarchitecture Simulation HPC Compiler Runtime GPU ×

Eidola addresses the problem of modeling irregular and transient inter-GPU communication traffic in distributed AI workloads, which existing tools fail to capture due to fine-grained synchronization and peer-to-peer writes. The method introduces a scalable gem5 extension that uses annotated timing profiles from real applications to emulate peer-to-peer GPU writes with cycle-level precision. Experimental evidence demonstrates Eidola's effectiveness by reproducing variability in fused kernel execution and confirming reductions in polling-related memory traffic via a SyncMon-inspired mechanism. This matters because Eidola provides a flexible platform for architectural exploration of interconnect bandwidth and latency in modern multi-GPU systems.

PDF

Harnessing Routing Foresight for Micro-step-level MoE load balancing in RL Post-training

Yuming Zhou, Haoyang Li, Sheng Lin, Yanfeng Zhao 2026-06-14
Language Model LLM Training GPU × Hardware

ForeMoE addresses expert load imbalance in Mixture-of-Experts (MoE) models during reinforcement learning (RL) post-training, where existing step-level statistics fail due to high-frequency micro-step fluctuations. The method exploits foreseeable routing information from the rollout stage to proactively guide load balancing, using a hierarchical planner to decompose the NP-hard problem and a transfer engine for overlapped expert transfer. Evaluations on 64 GPUs show up to a 1.45× speedup over state-of-the-art RL post-training systems. This matters because it enables efficient scaling of MoE LLMs under the unique workload dynamics of RL post-training, a dominant paradigm in current LLM development.

PDF

Maestro: Workload-Aware Cross-Cluster Scheduling for LLM-Based Multi-Agent Systems

Jinghao Wang, Xiao Zhou, Xiaoyang Sun, Yihui Zhang 2026-06-14
Workload × Scheduling LLM Agent

Maestro addresses the problem of high resource consumption and scheduling inefficiencies in deploying LLM-based multi-agent systems under strict GPU budgets. The method uses agent semantics to predict output length and memory usage, enabling hierarchical scheduling with dynamic model co-location, latency-aware routing, and workflow-aware prioritization. Experimental evidence shows Maestro reduces KV-reservation HBM by 67.2% and improves high-contention SLO attainment over EDF by 23.6 percentage points. This matters because it enables efficient, scalable deployment of complex multi-agent workflows in resource-constrained cloud environments.

PDF

Characterizing Software Aging in GPU-Based LLM Serving Systems

Domenico Cotroneo, Bojan Cukic 2026-06-14
GPU × LLM Serving

The paper addresses the problem of software aging in GPU-based LLM serving systems, which differ from traditional CPU-centric systems due to heterogeneous hardware and highly variable workloads. The method involves a 216-hour empirical campaign across six co-located deployments with identical stress, monitoring host, device, and client metrics and applying a statistical pipeline for autocorrelation and multiple testing. Experimental evidence shows statistically significant memory aging in all deployments, with leak rates strongly dependent on the serving runtime and configuration. This matters because it provides a reproducible framework bridging software aging and rejuvenation research with LLM serving, enabling future mitigation strategies.

PDF

GF-DiT: Scheduling Parallelism for Diffusion Transformer Serving

Xinwei Qiang, Yifan Hu, Shixuan Sun, Jing Yang 2026-06-14
Transformer GPU × Runtime Scheduling Parallelism Serving

The problem is that existing Diffusion Transformer (DiT) serving systems use static parallelism for each request, which is inefficient due to heterogeneity across requests, execution stages, and system conditions. GF-DiT introduces a policy-programmable runtime that dynamically adapts parallelism via an asynchronous execution abstraction and group-free collectives for low-overhead online GPU reallocation. Experimental evaluation in vLLM-Omni shows GF-DiT improves throughput by up to 6.01×, reduces mean latency by up to 95%, and lowers SLO violation rates by up to 90% compared to fixed-pipeline execution. This matters because it enables efficient, elastic DiT serving that treats GPU parallelism as a schedulable resource, significantly improving performance and service quality for image and video generation workloads.

PDF