# Resource Management for GPT-based Model Deployed on Clouds: Challenges, Solutions, and Future Directions

**Authors:** Yongkang Dang, Minxian Xu, Kejiang Ye  
**arXiv:** [2308.02970](https://arxiv.org/abs/2308.02970)  
**Date:** 2023-08-05  
**Summary page:** https://www.summarizepaper.com/en/paper/2308.02970v1/

## Key points

- Widespread adoption of large language models (LLMs) like GPT on cloud computing environments has increased resource demand
- Resource management in clouds faces significant challenges due to this surge in demand
- Authors aim to address these challenges by identifying unique characteristics of resource management for GPT-based models and proposing solutions
- Introduce a comprehensive resource management framework and specialized scheduling algorithms for GPT-based models
- Discuss future directions for improving resource management in GPT-based models
- Emphasize the importance of promoting sustainable development of GPT-based models
- Paper provides insightful analysis of challenges faced by resource management when deploying GPT-based models on clouds and offers valuable solutions

## Summary

The widespread adoption of large language models (LLMs), such as the Generative Pre-trained Transformer (GPT), deployed on cloud computing environments like Azure, has resulted in a significant increase in demand for resources. This surge in demand presents substantial challenges for resource management in clouds. In this paper, the authors aim to address these challenges by identifying the unique characteristics of resource management for GPT-based models and proposing potential solutions. They introduce a comprehensive resource management framework and specialized scheduling algorithms specifically designed for GPT-based models to facilitate effective resource management. Additionally, the authors discuss future directions for improving resource management in GPT-based models and emphasize the importance of promoting their sustainable development. Overall, this paper provides an insightful analysis of the challenges faced by resource management when deploying GPT-based models on clouds and offers valuable solutions to promote their sustainable development and application.