AliyunECSAlgos/LMID

日志与指标集成数据集（LMID）是一个大规模的多模态数据集，包含来自阿里巴巴云计算平台的真实世界数据，用于研究云计算故障预测的独特挑战。该数据集由100万条日志和来自180,000台物理机在4个月内的37维监控时间序列组成。LMID为研究多模态学习、极端类别不平衡、长序列问题、标签去噪等挑战提供了一个平台。它提供了一个基于服务器历史数据的二分类任务，判断服务器是否会在近期内发生故障，结合异常序列和时间序列作为样本。数据集以Parquet文件格式存储，分为训练集和测试集，存在显著的类别不平衡。

The Logs and Metrics Integration Dataset (LMID) is a large-scale multi-modal dataset consisting of real-world data from Alibabas cloud computing platform, designed for addressing the unique challenges of cloud computing failure prediction. Comprising 100 million logs and 37 dimensions of monitoring time series from 180,000 physical machines over a period of 4 months, LMID provides a platform to study challenges such as multi-modal learning, extreme class imbalance, long sequence issues, and label denoising. It offers a binary classification task to predict server failures based on historical data, combining exception sequences and time series. The dataset is stored in Parquet file format and is divided into training and testing sets, with significant class imbalance.