快速入门:浏览带有示例数据的指标顾问演示Quickstart: Explore the Metrics Advisor demo with example data


指标顾问演示为只读,无法加入你的数据。The Metrics Advisor demo is read-only, and you will not be able to onboard your data. 若要对数据使用该服务,请先创建指标顾问资源To use the service on your data, first create a Metrics Advisor resource.

使用本文快速开始浏览指标顾问中的主要功能。Use this article to quickly start exploring the key features in Metrics Advisor. 我们提供了一个演示网站,其中包含示例数据和预设配置,以便你熟悉功能完备的 Web 门户。We provide a demo site with sample data and preset configurations, for you to get familiar with the fully featured Web portal.

单击此链接转到演示网站。Click this link to go to the demo site.

查看可用的示例数据View the available sample data

登录到演示网站后,“数据馈送”页将会显示。After you sign into into the demo site, you will see a Data feed page. 数据馈送是时序数据的逻辑组,来自数据源查询。a data feed is a logical group of time series data, which is queried from your data source. 有关指标顾问中使用的术语和概念的详细信息,请参阅术语表For more information on terms and concepts used in Metrics Advisor, see the Glossary.

此处列出了多个数据馈送,它们从不同类型的数据源(例如 Azure SQL 数据库或 Azure 表)引入。There are several data feeds listed, which are ingested from different types of data sources, such as Azure SQL database or Azure Table. 每个数据馈送都使用略微不同的配置设置来连接到关联的数据存储。Each uses slightly different configuration settings to connect to the associated data stores.


浏览数据馈送配置Explore the data feed configurations

单击“Sample - Cost/Revenue - City/Category”数据馈送。Click on the Sample - Cost/Revenue - City/Category data feed. 你将看到馈送的多个细节部分:You'll see several sections of details for the feed:

  • 数据馈送名称和引入状态。Data feed name and ingestion status.
  • 从数据源查询的指标列表。A list of metrics queried from the data source. 例如“cost”和“revenue” 。For example, cost and revenue .
  • 数据馈送变为不可用时的警报历史记录。Alert history for when the data feed becomes unavailable.
  • 数据馈送更新时间的日志。Logs of when the data feed was updated.
  • 数据馈送信息和设置。Data feed information and settings.


查看时序可视化效果和配置View time series visualizations and configurations

单击进入“Sample - Cost/Revenue - City/Category”数据馈送中的“cost”指标 。Click into the cost metric in the Sample - Cost/Revenue - City/Category data feed. 你将看到根据维度切片的关联时序,以及根据历史指标数据显示的可视化效果。You'll see the associated time series sliced by dimensions, with visualizations according to the historical metric data. 指标数据周围的蓝带表示指标顾问的机器学习模型中的预期值范围。The blue band around the metric data represents the expected value range from Metrics Advisor's machine learning models. 位于此带之外的点将以红点标记,表示检测到异常。Points that fall outside of this band will be marked as red dots, which are detected anomalies.


可以通过在指标详细信息页的左侧优化“检测配置”来配置异常情况检测。The anomaly detection is configurable by tuning the detecting configurations on the left side of metric details page. 有多种异常情况检测方法可用,你可以将这些方法加以组合。Multiple anomaly detection methods are available and you can combine them. 还可以尝试不同的敏感度、检测方向和其他配置。You can additionally try different sensitivities, detecting directions, and other configurations. 通过“检测配置”底部的”高级配置”链接,可以创建更复杂的自定义检测设置,这些设置可用于组或单个序列 。The Advanced configuration link at the bottom of detecting configurations lets you create more complex and customized detection settings, which can be used on groups or individual series.

还可以通过向检测算法提供反馈来优化异常情况检测。You can also tune anomaly detection by providing feedback to the detection algorithm. 单击进入异常,并使用“添加反馈”面板配置其异常状态、季节性和更改点状态。Click into an anomaly, and use the Add feedback panel to configure its anomaly status, seasonality, and change point status. 此反馈将融入将来点的检测中。This feedback will be incorporated in the detection for future points.

面板中的底部有一个“前往事件中心”链接,该链接会将你定向到事件分析页并分析事件的根本原因 。At the bottom of the Add feedback panel, there's a link of To incident hub , which will direct you to incident analysis page and analyze root cause of the incident.


浏览异常情况检测结果并执行根本原因分析Explore anomaly detection results and perform root cause analysis

从异常中单击“前往事件中心”链接时,你将看到一个事件分析页,其中包含有关事件的诊断见解,如严重性、涉及的异常数量以及开始/结束时间。When you click the To incident hub link from an anomaly, you will see an incident analysis page, containing diagnostics insights about the incident, such as Severity, the number of anomalies involved, and start/end time. “根本原因”部分通过分析事件树自动显示建议,并考虑到偏差、分布和对父异常的贡献,而这些可能是事件的根本原因。The Root cause section displays automated advice by analyzing the incident tree, taking into account: deviation, distribution and contribution to parent anomalies, which may be the root cause of the incident.

“诊断”部分显示事件树以及用于诊断事件的多个选项卡。The Diagnostics section shows a tree of the incident, along with several tabs for diagnosing the incident.


通过查明事件的根本原因,你可以在情况恶化之前采取措施并缓解问题。By pinpointing root cause of the incident, you can take take action and mitigate the issue before the situation gets worse. 还可以单击提供的其他诊断功能来浏览更多见解。You can also explore more insights by clicking through the other diagnostic features provided.

后续步骤Next steps