Daily | Yixun Hong

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

ITME: Inference Tiered Memory Expansion with Disaggregated CXL-Hybrid Memories

Hakbeom Jang, Younghoon Min, Sunwoong Kim, Taeyoung Ahn 2026-06-14

Inference

ITME addresses the problem of scaling shared context infrastructure for TB-scale LLM inference workloads beyond individual server capacity. The method leverages CXL-hybrid memory to provide massive, byte-addressable remote memory expansion, simplifying the software stack by eliminating complex software-level optimization. Experimental evidence from production-grade SK Hynix CMM and PCIe Gen5 NVMe SSDs, along with an FPGA prototype, shows up to a 35.7% throughput improvement over conventional CPU-offloading. This matters because ITME enables cost-efficient scaling of shared context layers for agentic and long-context LLMs by proactively managing data movement across the memory-storage hierarchy.

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

Work Stealing for the 2D-Mesh Topology of Satellite Constellations in Low Earth Orbit

Mia Reitz, Dorian Chenet, Jonas Posner 2026-06-14

High Performance Computing HPC Compiler Runtime

The problem is that existing Asynchronous Many-Task (AMT) runtimes assume a fully connected network with low, uniform latency, which is invalid for satellite constellations in Low Earth Orbit (LEO) that communicate via a sparse mesh topology. The method proposes a neighbor-only work stealing strategy where workers steal exclusively from directly connected neighbors to avoid multi-hop communication. Experimental evidence on an HPC cluster with an emulated mesh shows the neighbor-only strategy performs within ~2.2% of global stealing on both balanced and irregular workloads, and an analytical model indicates a growing latency advantage with constellation size. This matters because it demonstrates that neighbor-only stealing can match global stealing performance in emulated settings, suggesting it is a viable and potentially preferable approach for adapting AMT to Space Edge Computing (SEC) at scale.

PDF