No Request Left Behind: Tackling Heterogeneity in Long-Context LLM Inference with Medha

TL;DR

Medha is a novel serving system for large language models that employs preemptive scheduling and parallelism strategies to handle heterogeneous workloads efficiently, significantly improving throughput and latency.

Contribution

Medha introduces a comprehensive preemptive scheduling framework with mechanisms like Adaptive Chunking, Stream Pipeline Parallel, and KV-Cache Parallelism, orchestrated by the LARS scheduler, to mitigate convoy effects in LLM inference.

Findings

5.7x throughput improvement under heterogeneous workload

30x reduction in median latency

174x reduction in 99th percentile latency

Abstract

Deploying million-token Large Language Models (LLMs) is challenging because production workloads are highly heterogeneous, mixing short queries and long documents. This heterogeneity, combined with the quadratic complexity of attention, creates severe convoy effects where long-running requests stall short, interactive ones, degrading system responsiveness. We present Medha, a serving system that eliminates these convoys by introducing fine-grained, preemptive scheduling to LLM inference. Medha makes preemption practical with a co-designed set of mechanisms -- including Adaptive Chunking and Stream Pipeline Parallel that overcome the perceived inefficiencies and scaling challenges of chunking. Additionally, we present a new parallelism strategy KV-Cache Parallelism to reduce the decode latency and afford interactivity despite very long context. These mechanisms are orchestrated by a Length-Aware Relative Slack (LARS) scheduler, a deadline and heterogeneity-aware scheduling policy that prevents both the convoy effect and the starvation that plagues simpler policies. Under a heterogeneous workload, Medha improves throughput by 5.7x while reducing median and 99th percentile latency by 30x and 174x, respectively, compared to state-of-the-art non-preemptive systems.