Alert Notification

TingYun Network provides alert notification functionality. Users can flexibly formulate alert strategies based on existing task requirements, supporting three alert methods: Performance Alerts, Error Alerts, and Smart Alerts. The system can trigger based on performance, errors, or availability separately. The alert system will send notifications via email and SMS. Smart alerts are based on machine learning principles, eliminating complex threshold configuration work, and through a long-term learning process, gradually aligning with users' alert cognition based on user feedback.

Performance Alerts

Set various performance metrics and availability alerts for monitoring tasks. When the performance values or availability of these performance metric items for monitoring tasks exceed the set thresholds, the system will send alerts via email and SMS and generate alert logs.

Performance alert functionality provides Static Alerts and Dynamic Alerts. System administrators can flexibly configure alerts based on their website conditions.

Static Alerts directly set a static performance value and a static availability value, setting alert trigger conditions that trigger when all conditions are met simultaneously or when any one condition is met.

Since some website content or other elements change constantly, static alerts may not be flexible enough. The biggest difference with Dynamic Alerts is that they can change dynamically based on an alert foundation. System administrators can specify a calculation period, during which alerts can flexibly change.

Error Alerts

System administrators can set alerts for when errors occur in monitoring tasks, selecting one or more errors as alert trigger conditions. When these errors occur in monitoring tasks, alerts are triggered and alert logs are generated.

Smart Alerts

Smart alerts employ anomaly detection and intelligent alert algorithms based on time series prediction.

Time Series Prediction-Based Anomaly Detection Algorithm: Statistical methods based on data models are the most widely used anomaly detection techniques. The basic principle is: perform statistical modeling on training datasets (usually normal examples). If a data sample does not conform to the trained stochastic model, it is identified as an anomaly sample.

Machine Learning-Based Intelligent Alert Algorithm: Through time series prediction-based anomaly detection algorithms, we can enable the alert system to automatically and relatively accurately complete fault detection and alerting tasks, thereby greatly reducing the burdensome manual setting of various performance metric alert thresholds and manual management tasks, achieving the goal of automated and intelligent fault detection.

Alert Logs

Specific information when alerts occur can be viewed through alert logs, including Performance Alert Logs, Error Alert Logs, and Smart Alert Logs.

In alert details, you can view Alert Overview, Alert Timeline, Performance Scatter Plot, and Probe access records.

Alert Push

Alert push for Network products is divided into two types: HTTP Push and UDP Push. More and more customers are beginning to use unified information access platforms to receive information pushed by third-party service vendors, such as alert information. Network products have implemented templated alert push methods, where standard alert fields can be concatenated into required formats according to customer requirements, meeting customers' diverse needs for alert information.

Performance Alerts#

Error Alerts#

Smart Alerts#

Alert Logs#

Alert Push#

Performance Alerts

Error Alerts

Smart Alerts

Alert Logs

Alert Push