一句话：提出原型工具 ImpAPTr，用剪枝树分析多维属性组合，以及时定位服务调用成功率下降的根因。
本文关注 APM 系统中服务调用 DSR（Declining Success Rate）异常的根因定位问题。由于每个调用请求包含 City、ISP、Platform 等多维属性，且属性可有多个取值，异常可能由单一属性或属性组合引起，根因复杂且难以快速识别。作者提出原型工具 ImpAPTr（Impact Analysis based on Pruning Tree），通过剪枝树分析多维属性的影响，找出可作为线索的属性组合，从而及时挖掘 DSR 异常的根因。该工具在美团这一大型在线服务提供方中评估，准确率优于同领域若干已有工具。
As a common IT infrastructure, APM (Application Performance Management) systems have been widely adopted to monitor call requests to an on-line service. Usually, each request may contain multi-dimensional attributes (e.g., City, ISP, Platform, etc.), which may become the reason for a certain anomaly regarding DSR (Declining Success Rate) of service calls either solely or as a combination. Moreover, each attribute may also have multiple values (e.g., ISP could be T-Mobile, Vodafone, CMCC, etc.), rendering intricate root causes and huge challenges to identify the root causes. In this paper, we propose a prototype tool, ImpAPTr (Impact Analysis based on Pruning Tree), to identify the combination of dimensional attributes as the clues to dig out the root causes of anomalies regarding DSR of a service call in a timely manner. ImpAPTr has been evaluated in MeiTuan, one of the biggest on-line service providers. Performance regarding the accuracy outperforms several previous tools in the same field.
DOI 原文 ·
@inproceedings{paperbot6022,
title = {ImpAPTr},
author = {Hao Wang and Guoping Rong and Yangchen Xu and Yong You},
booktitle = {ASE},
year = {2020},
doi = {10.1145/3324884.3415301}
}