Oneiros: KV Cache Optimization through Parameter Remapping for Multi-tenant LLM Serving

TL;DR

Oneiros introduces a parameter remapping technique to optimize KV cache management in multi-tenant LLM serving, significantly reducing latency and improving throughput by avoiding costly cache swapping.

Contribution

The paper presents a novel parameter remapping approach that eliminates KV cache swapping, enabling more efficient multi-tenant LLM inference on modern hardware.

Findings

Reduces tail latency by up to 82.5%.

Increases throughput by up to 86.7%.

Outperforms state-of-the-art solutions significantly.

Abstract

KV cache accelerates LLM inference by avoiding redundant computation, at the expense of memory. To support larger KV caches, prior work extends GPU memory with CPU memory via CPU-offloading. This involves swapping KV cache between GPU and CPU memory. However, because the cache updates dynamically, such swapping incurs high CPU memory traffic. We make a key observation that model parameters remain constant during runtime, unlike the dynamically updated KV cache. Building on this, we introduce Oneiros, which avoids KV cache swapping by remapping, and thereby repurposing, the memory allocated to model parameters for KV cache. This parameter remapping is especially beneficial in multi-tenant environments, where the memory used for the parameters of the inactive models can be more aggressively reclaimed. Exploiting the high CPU-GPU bandwidth offered by the modern hardware, such as the NVIDIA Grace Hopper Superchip, we show that Oneiros significantly outperforms state-of-the-art solutions, achieving a reduction of 44.8%-82.5% in tail time-between-token latency, 20.7%-99.3% in tail time-to-first-token latency, and 6.6%-86.7% higher throughput compared to vLLM. Source code of Oneiros is available at https://github.com/UT-SysML/Oneiros/.