29515

NEO: Saving GPU Memory Crisis with CPU Offloading for Online LLM Inference

Xuanlin Jiang, Yang Zhou, Shiyi Cao, Ion Stoica, Minlan Yu
Harvard University
arXiv:2411.01142 [cs.DC], (2 Nov 2024)

@misc{jiang2024neosavinggpumemory,

   title={NEO: Saving GPU Memory Crisis with CPU Offloading for Online LLM Inference},

   author={Xuanlin Jiang and Yang Zhou and Shiyi Cao and Ion Stoica and Minlan Yu},

   year={2024},

   eprint={2411.01142},

   archivePrefix={arXiv},

   primaryClass={cs.DC},

   url={https://arxiv.org/abs/2411.01142}

}

Download Download (PDF)   View View   Source Source   

528

views

Online LLM inference powers many exciting applications such as intelligent chatbots and autonomous agents. Modern LLM inference engines widely rely on request batching to improve inference throughput, aiming to make it cost-efficient when running on expensive GPU accelerators. However, the limited GPU memory has largely limited the batch size achieved in practice, leaving significant GPU compute resources wasted. We present NEO, an online LLM inference system that offloads part of attention compute and KV cache states from the GPU to the local host CPU, effectively increasing the GPU batch size and thus inference throughput. To this end, NEO proposes asymmetric GPU-CPU pipelining and load-aware scheduling to balance GPU and CPU loads and fully utilize their compute and memory resources. We evaluate NEO on a wide range of workloads (i.e., code generation, text summarization), GPUs (i.e., T4, A10G, H100), and LLM models (i.e., 7B, 8B, 70B). NEO achieves up to 7.5x, 26%, and 14% higher throughput compared to GPU-only approach on T4, A10G, and H100 GPUs, respectively, while maintaining the same latency; with more powerful CPUs, NEO achieves up to 79.3% throughput gain on A10G GPU.
No votes yet.
Please wait...

* * *

* * *

HGPU group © 2010-2024 hgpu.org

All rights belong to the respective authors

Contact us: