Balancing and Beyond: Communication-Centric Optimizations in Expert Parallelism

Fine-grained Computation-Communication Overlap via Tile-level Signaling and Scheduling for Mixture-of-Experts

Mixture-of-Experts (MoE) architectures increase model capacity without proportionally increasing computation cost and have become a key building block for scaling large language models (LLMs) to trillion-parameter regimes. Efficient deployment of these MoE models relies on distributed execution across multiple GPUs, where each MoE layer involves two all-to-all communications: dispatching tokens to expert ranks and returning the expert outputs to their source ranks. Conventional MoE implementations launch this return all-to-all after expert compute completes, exposing communication latency on the critical path and reducing GPU utilization. We present a fine-grained approach that overlaps expert compute with the second all-to-all via tile-level signaling and scheduling. Our producer-consumer co-design combines: (1) a persistent per-rank computation kernel (producer) that covers all local experts on the rank to eliminate repeated kernel launch overhead and prioritizes remote-critical tiles, and (2) a persistent communication kernel (consumer) on a small dedicated partition of streaming multiprocessors (SMs) that issues segment-granular transfers as tiles become ready. Our co-design avoids intrusive changes to the underlying computation operators or communication primitives, making it practical for improving distributed MoE execution efficiency on multi-GPU systems. On a 4-A100 GPU platform, evaluated on three MoE models against four state-of-the-art MoE systems, our approach achieves up to 2.64x end-to-end speedup and 2.74x MoE-layer speedup. Compared with a conventional non-overlap baseline, our approach consistently improves both operator- and MoE-layer-level performance across varying GEMM shapes, router modes, and a broad range of producer/consumer SM partitions, while preserving correctness.

OrionInfer: Low-Overhead Parallelism Switching and Live Migration for Efficient LLM Serving

Existing Large Language Model (LLM) inference systems often rely on static model placement and scheduling policies, which struggle to handle heterogeneous and dynamic real-world workloads. The key challenge is to adapt serving strategies to workload fluctuations while keeping reconfiguration overhead minimal. In this paper, we present OrionInfer, an adaptive LLM serving system that aligns inference strategies with real-time demand. OrionInfer introduces three key techniques: (1) runtime switching between data parallelism and tensor parallelism with negligible overhead; (2) an efficient inference pipeline that preserves batching efficiency during parallelism transitions; and (3) live-migration-based load balancing to alleviate memory pressure and improve resource utilization. Evaluations across multiple model scales show that OrionInfer delivers robust performance under diverse serving scenarios. In end-to-end serving, it reduces average TTFT by up to 25% over DP-priority configurations under low loads and lowers P99 tail latency by 50%--90% over TP-priority configurations under most high-traffic settings. In disaggregated prefill serving, OrionInfer improves prefill completion time (PCT) SLO attainment by up to 16.5 percentage points over DP-priority static baselines and reduces P99 PCT by up to 74.7% over TP-priority static baselines. Compared with dynamic baseline, OrionInfer provides better tail-latency stability, reducing P99 PCT by 38.6%--40.8% while avoiding the extra memory footprint.

Rethinking LLM Deployment for Intent-Based Serving

Large Language Models (LLMs) expose a large deployment configuration space spanning parallelism and compression techniques, with each configuration introducing different tradeoffs in latency, memory consumption, cost, and output quality. Existing systems either rely on expensive profiling across deployment configurations or inefficiently utilize fragmented GPU resources in multi-tenant clusters. We present MaverIQ, an intent-based LLM inference serving system that automatically maps user intents to deployment configurations while minimizing operational cost for the provider. To reduce profiling overheads, MaverIQ introduces lightweight LLM fingerprints and analytical models that extrapolate latency and memory footprint from only a few observations. To efficiently utilize fragmented GPU resources, MaverIQ leverages our observation that, unlike training, unevenly distributing LLM layers across GPUs has little impact on inference latency. Our evaluation shows that MaverIQ reduces profiling cost by 7-15× compared to state-of-the-art baselines and reduces operational cost by 3.8-8.3× across diverse LLMs, traces, and loads while effectively meeting user intents. Our code is available at https://github.com/UT-SysML/MaverIQ.

Reproducible Evaluation of MoE Expert Caching: Replay Semantics, Workload Contamination, and Operating Regimes

With a trace-driven, event-atomic simulator over three MoE models, a large offline-optimal gap substantially overstates the gains recovered by representative lightweight causal mechanisms.

LLM4LLM: Bridging Kernel Benchmarks and Real Deployment via Closed-Loop Agentic Optimization

Large language models have become increasingly capable agents for low-level code and kernel optimization, but isolated kernel benchmarks provide only a proxy for the deployment behavior that matters in language-model inference. We identify a benchmark-to-deployment gap: candidate kernels that appear correct and fast in standalone harnesses can exhibit different performance, safety, or phase behavior after integration into a real inference workload. We introduce LLM4LLM, a deployment-aware closed-loop optimization framework that starts from a target inference script, extracts phase-aware optimization tasks, searches with an experience-guided episodic agent, and accepts patches through in-model validation. Across ten language-model inference workloads on A100 and H100 GPUs, LLM4LLM improves end-to-end latency for every evaluated model, achieving 3.91$\times$/6.98$\times$ geometric-mean speedups on A100/H100; as supporting kernel-level evidence, it also attains up to 2.745$\times$ GeoMean speedup on KernelBench Level 2.

Dynamic Pipeline Inference Optimization for LLMs: Load-Aware Partitioning, Updating, and Migration