NetLLM: Adapting Large Language Models for Networking

TL;DR

This paper tackles the high engineering cost and limited generalization of DL-based networking solutions by proposing NetLLM, a framework that adapts large language models (LLMs) to networking tasks. NetLLM introduces a multimodal encoder to convert diverse input modalities into token-like representations, a networking head to generate valid task outputs in a single inference, and a data-driven low-rank networking adaptation (DD-LRNA) to efficiently acquire domain-specific knowledge without full fine-tuning. Evaluated on viewport prediction, adaptive bitrate streaming, and cluster job scheduling, NetLLM-based LLMs consistently outperform state-of-the-art baselines and exhibit stronger generalization to unseen environments, while reducing adaptation costs (e.g., memory and training time) and enabling reuse of a single foundation model across tasks. The work demonstrates that LLMs can serve as foundation models for networking, offering a more sustainable design paradigm with potential transferability to other domains and inspiring further research in efficient multimodal, domain-aware AI for systems.

Abstract

Many networking tasks now employ deep learning (DL) to solve complex prediction and optimization problems. However, current design philosophy of DL-based algorithms entails intensive engineering overhead due to the manual design of deep neural networks (DNNs) for different networking tasks. Besides, DNNs tend to achieve poor generalization performance on unseen data distributions/environments. Motivated by the recent success of large language models (LLMs), this work studies the LLM adaptation for networking to explore a more sustainable design philosophy. With the powerful pre-trained knowledge, the LLM is promising to serve as the foundation model to achieve "one model for all tasks" with even better performance and stronger generalization. In pursuit of this vision, we present NetLLM, the first framework that provides a coherent design to harness the powerful capabilities of LLMs with low efforts to solve networking problems. Specifically, NetLLM empowers the LLM to effectively process multimodal data in networking and efficiently generate task-specific answers. Besides, NetLLM drastically reduces the costs of fine-tuning the LLM to acquire domain knowledge for networking. Across three networking-related use cases - viewport prediction, adaptive bitrate streaming and cluster job scheduling, we showcase that the NetLLM-adapted LLM significantly outperforms state-of-the-art algorithms.

Figures (17)

• Figure 1: Illustration of the token-based answer generation mechanism of LLMs.
• Figure 2: Illustration of the ineffectiveness for some natural alternatives with VP task as the example. Left: Prompt learning liu2023premin2023recent that transforms data into textual prompts achieves sub-optimal performance, while NetLLM with a multimodal encoder to encode task input data effectively outperforms baseline. Middle, Right: Token-based prediction with LM head fails to guarantee valid answers and produce stale responses, while NetLLM efficiently addresses these issues with the networking head module.
• Figure 3: Using standard RL techniques van2016deepschulman2017proximal to adapt LLM for RL-based decision-making tasks (ABR and CJS) incurs high training time due to the active environment interaction for experience collection. NetLLM eliminates this time-consuming process by designing an efficient data-driven adaptation pipeline in the DD-LRNA scheme.
• Figure 4: Illustration of the high adaptation costs of full-parameter fine-tune touvron2023llamachung2022scaling on the VP task. The DD-LRNA scheme of NetLLM efficiently reduces the costs by introducing a set of small trainable low-rank matrices.
• Figure 5: NetLLM consists of three core components: multimodal encoder to encode task inputs, networking head to generate task-specific answers and data-driven low-rank networking adaptation to efficiently learn domain knowledge for networking. The framework is illustrated with three tasks: VP, ABR and CJS, but all ideas can be easily applied to other networking tasks.