AutoScale real-world traces NLAR T4 and T5
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/pacslab/serverless-ml-serving
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了模拟服务器无感知计算实验中到达模式的真实世界追踪数据。此外，该数据集被用于匹配集群的容量并模仿不同的负载强度。为了实验目的，数据集已被缩放到每秒最大100个请求的到达率。任务是对不同负载强度下Mlproxy效率的评估。
This dataset contains real-world trace data on arrival patterns in serverless computing simulation experiments. Additionally, it is used to match cluster capacity and emulate varying load intensities. For experimental purposes, the dataset has been scaled to a maximum arrival rate of 100 requests per second. The core task of this dataset is to evaluate the efficiency of Mlproxy under different load intensities.
