FAMOS公开数据集

The Train-Ticket and Mall datasets were constructed for the fault diagnosis task. Both datasets contain three modalities of observable data: traces, logs, and metrics. We constructed these datasets through fault injection on both a benchmark microservice system and a real industrial system. Our objective is to create two multimodal datasets with a rich variety of fault types, making them suitable for the fault diagnosis task.

Train-Ticket与Mall数据集（Train-Ticket and Mall datasets）专为故障诊断任务构建。两类数据集均包含三种可观测数据模态：链路追踪数据（traces）、日志（logs）与指标（metrics）。本研究通过在基准微服务系统与真实工业系统中开展故障注入操作，构建了上述两类数据集。本次构建的核心目标是打造覆盖丰富故障类型的多模态数据集，使其适配故障诊断任务需求。

搜集汇总

背景与挑战

背景概述

FAMOS公开数据集包含Train-Ticket和Mall两个微服务系统的故障诊断数据，提供追踪、日志和指标三种数据模态，涵盖16种故障类型。数据集通过故障注入生成，并包含详细的故障标签和时间信息，适用于微服务故障诊断研究。