AIOPS论文的综述
如何挑选论文，如何选取keywords
搜索的3个论文库，
We select three online search databases that are appropriate for the scope of investigation: IEEE Xplore, ACM Digital Library and arXiv.
对于挑选出的论文进行分类，
分类标准是，target components and data sources.
Target components，ops什么？
indicate a high-level piece of software or hard- ware in an IT system that the document tries to enhance (e.g. hard drive for hard disk failure prediction).
We group components in five high-level categories: code, application, hardware, network and datacenter.
Data sources，输入什么？
provide an indication of the input information of the algorithm (such as logs, metrics, or execution traces).
Data sources are categorized in source code, testing resources, system metrics, key performance indicators (KPIs), network traffic, topology, incident reports, logs and traces.
这里进行了分类，20多篇的paper，集中的领域主要在
- 硬件和系统的失败预测
- 异常检测
- 异常定位和根因分析
将AIOps分为两个领域，
Failure Management，这个占据61%，其中细分主要是，Failure detection，RCA
Resource Provisioning，主要是资源调度（resource consolidation, scheduling）和负载预测（ workload prediction）
从时间线的角度上来看，2020前后Failure Detection的论文大幅增长
