DUAL-BLADE: Dual-Path NVMe-Direct KV-Cache Offloading for Edge LLM Inference

TL;DR

DUAL-BLADE is a dual-path KV-cache offloading framework for edge LLM inference that dynamically balances memory and NVMe storage to reduce latency and improve throughput.

Contribution

It introduces a runtime adaptive framework that assigns KV tensors to page-cache or NVMe paths, enabling low-overhead direct storage access and overlapping I/O with GPU processing.

Findings

Reduces prefill latency by up to 33.1%

Decreases decode latency by up to 42.4%

Improves SSD utilization by 2.2x

Abstract

The increasing deployment of Large Language Model (LLM) inference on edge AI systems demands efficient execution under tight memory budgets. A key challenge arises from Key-Value (KV) caches, which often exceed available device memory. Although NVMe-based offloading offers scalable capacity, existing file-based designs rely heavily on the kernel page cache, leading to cache thrashing, unpredictable latency, and high software overhead under memory pressure. We present DUAL-BLADE, a dual-path KV residency framework that dynamically assigns KV tensors to either a page-cache path or an NVMe-direct path based on runtime memory availability. The NVMe-direct path bypasses the filesystem by mapping KV tensors to contiguous logical block address (LBA) regions, enabling low-overhead direct storage access. DUAL-BLADE further incorporates adaptive pipeline parallelism to overlap storage I/O with GPU DMA, improving inference throughput. Our evaluation shows that DUAL-BLADE substantially mitigates I/O bottlenecks, reducing prefill and decode latency by up to 33.1% and 42.4%, respectively, while improving SSD utilization by 2.2x across diverse memory budgets.