@inproceedings{yang2025Prism, title = {{{GPU-disaggregated}} serving for deep learning recommendation models at scale}, author = {Yang, Lingyun and Wang, Yongchen and Yu, Yinghao and Weng, Qizhen and Dong, Jianbo and Liu, Kan and Zhang, Chi and Zi, Yanyi and Li, Hao and Zhang, Zechao and Wang, Nan and Dong, Yu and Zheng, Menglei and Xi, Lanlan and Lu, Xiaowei and Ye, Liang and Yang, Guodong and Fu, Binzhang and Lan, Tao and Zhang, Liping and Qu, Lin and Wang, Wei}, booktitle = {Proc. USENIX NSDI}, year = {2025} }