Cloud Monitor:What Is Cloud Monitor 2.0

Term

Description

Workspace

A workspace is an abstraction layer in Cloud Monitor 2.0 that groups resources with unified management and data isolation. The selected region stores workspace data and configuration. Each workspace contains its own resources—cloud services, infrastructure, server-side and frontend applications, and middleware—isolated from other workspaces.

• The workspace region matches the data source region. Data stays local and is also stored in EntityStore.

• WorkSpaceName is globally unique—like ProjectName. You can enter it when creating the workspace. If left blank, the system generates one. After creation, you cannot change the WorkSpaceName. It serves as the unique identifier in API calls.

• WorkSpaceDisplayName is editable and allows duplicates.

• Default workspace: named default-cms-{uid}-{regionId}. The linked SLS project follows a similar format, such as default-cms-{uid}-{regionId}, but may include extra random characters.

• The first workspace created in a region is always the default workspace.

App

An app is a lightweight carrier for reading and writing data sources within a workspace, representing domain-specific observability for a given scenario. Key traits:

• Apps can be enabled or disabled per user (RAM user)—including whether they appear in the navigation pane on the left. Disabled apps can be re-enabled anytime.

• Apps are stateless. All apps are available in every workspace, but data visibility depends on user permissions.

• An app can access multiple data sources. Switching between them reveals IaaS-layer relationships.

• Integration Center defines app initialization and data source creation.

• Apps can embed other apps. For example, Explorer and Alerting are embedded in many apps and also run standalone.

Entity

An entity is an observable object—such as a container cluster or an ECS instance.

Model (Umodel)

UModel is a specification for defining observability data models. It defines models for logs, metrics, traces, entities, and their relationships—enabling unified definition and management of observability data.

Statistical period (Period)

The statistical period (Period) is the time granularity at which monitoring data is aggregated. It determines how much raw data a single data point covers.

• Statistical values such as Maximum, Minimum, and Average are aggregated from the raw data collected within the selected statistical period, not instantaneous samples. For example, when the statistical period is 300 seconds, Average is the average of all raw data in those 300 seconds. Maximum and Minimum are the highest and lowest values in the same period.

• The available statistical periods depend on the metric. Common values are 300 seconds, 900 seconds, and 3600 seconds, and the minimum supported value is 60 seconds.

• A shorter statistical period means more frequent collection and aggregation, denser data points, and finer-grained fluctuations, but higher resource consumption. A longer statistical period means less frequent collection and lower resource consumption, with smoother curves in which short-lived fluctuations may be averaged out.

Features

Description

Full-stack data collection and monitoring

• Infrastructure monitoring

  • Cloud service monitoring: Collect real-time performance metrics—such as CPU, memory, disk, and network traffic—for ECS, RDS, SLB, container services (ACK/ASK), and Kubernetes pods.

  • Network performance monitoring: Track network latency, packet loss rate, DNS resolution, and TCP/UDP connection status.

• Application performance monitoring (APM)

  • Distributed tracing: Trace microservice call chains. Supports Java, Python, Go, and other major languages. Shows interface latency, error rate, and dependency topology.

  • Code-level diagnostics: Locate application performance bottlenecks using thread profiling, slow SQL detection, and stack trace analysis.

• Log monitoring

  • Log collection and storage: Collect logs from servers, containers, and Function Compute (FC). Compatible with Log4j, Logback, and other logging frameworks.

  • Real-time log analysis: Run SQL queries, set keyword alerts, and use log clustering—such as aggregating error logs.

Intelligent analysis and diagnostics

• Anomaly detection and alerting

  • Dynamic threshold alerting: Use machine learning to learn historical metric patterns and spot anomalies—such as sudden CPU spikes.

  • Multi-condition alerting: Trigger alerts based on cross-metric conditions—such as “CPU > 90% AND packet loss rate > 5%”.

• Root cause analysis (RCA)

  • Smart association analysis: Automatically link abnormal metrics, error logs, and trace data to generate root cause reports—such as an API timeout triggering a downstream service avalanche.

  • Time-series data rollback: Compare historical data to quickly find when an anomaly started and its impact scope.

Visualization and reporting

• Custom monitoring dashboards

  • Drag-and-drop dashboard: Use line charts, column charts, topology graphs, and other visual components to flexibly display key metrics across resources and services.

  • Scenario-based template library: Prebuilt templates for e-commerce sales promotions, container clusters, and database performance. Generate business-wide views with one click.

• Business dashboards

  • Real-time data projection: Show core business metrics—such as order volume and payment success rate—in full-screen mode. Designed for O&M war rooms.

  • Multi-tenant view isolation: Assign data viewing permissions by team or line-of-business to keep data secure.

Alerting and notification management

• Multi-channel alert delivery

  • Notification channels: Send alerts via DingTalk, WeCom, text message, email, and webhook. Support time-based silence—such as notifying only on-call staff outside business hours.

  • Alert escalation policy: Set tiered alerts—such as “notice → critical → fatal”. Escalate notifications automatically if no response arrives on time.

• Alert lifecycle management

  • Alert history and statistics: Record alert status—such as confirmed or recovered—and generate MTTR (mean time to repair) reports.

  • Integration with O&M tools: Trigger ticket systems—such as DingTalk Yida—or automated O&M scripts—such as restarting a service—when alerts fire.

Openness and integration

• Seamless ecosystem integration

  • Alibaba Cloud service integration: Deeply links with SLS, ARMS, and Alibaba Cloud DevOps. Enables automatic data association—such as jumping from a log query directly to an abnormal trace.

  • Third-party tool compatibility: Supports open-source protocols like Prometheus, OpenTelemetry, and Telegraf. Works with Grafana visualization and Jenkins continuous integration.

  • Infrastructure as code (IaC): Terraform supports Cloud Monitor 2.0. You can use Terraform to manage metric stores and refresh monitoring metrics. For more information, see Terraform alicloud cms_addon_release.

• API and SDK support

  • OpenAPI management: Automate monitoring tasks—such as batch-creating alert rules or exporting monitoring data—using APIs.

  • Custom metric reporting: Use SDKs to report business metrics—such as order volume or activity PV/UV—to extend monitoring coverage.

  • STAROps (formerly AIOPS) API access: STAROps is deployed at region-group level. Requests from regions in China are served by the Beijing region group, and requests from regions outside China are served by the Singapore region group. When you call STAROps APIs from within China, set the Endpoint to cms.cn-beijing.aliyuncs.com (cn-beijing). A call from a region that is not covered by these region groups, such as Shenzhen, returns a service-unavailable error.

    The STAROps public preview has ended. Calls to STAROps APIs, including CreateThread, consume STAROps quota and are billed based on usage once the monthly free quota is exceeded.

Security and high availability

• Data security

  • End-to-end encryption: Encrypt monitoring data during transmission (HTTPS) and storage (encrypted storage).

  • Permission control: Use RAM roles for fine-grained permission management—such as read-only access or alert configuration permissions.

  • Security and compliance: Meets international and domestic security standards to ensure safe and reliable data transmission and storage.

• Service reliability

  • Global distributed collection points: Monitoring nodes cover major regions worldwide to avoid data loss due to network jitter.

  • Data redundancy: Store monitoring data across multiple replicas to ensure recoverability.

Cost optimization

• Resource usage analysis

  • Idle resource identification: Auto-flag long-term low-load ECS instances and unbound EIPs. Generate release recommendations.

  • Cost allocation reports: Track cloud resource consumption by project, department, or tag. Support cost allocation and budget control.

• Adaptive data sampling

  • Adjust collection frequency as needed: Reduce frequency for non-critical metrics—such as from 1 minute to 5 minutes—to cut storage costs.

Cross-region unified management

Monitor and manage resources across multiple regions from one place. Simplify O&M workflows.

Scenario

Description

Advantages

Scenario 1: Unified full-stack monitoring and real-time observability graph

Enterprises need to monitor physical servers, containers, microservices, and databases across hybrid cloud environments. Cloud Monitor 2.0 collects metrics, traces, logs, and events in one place and builds an end-to-end observability graph for cross-resource, cross-service visibility.

• Multi-source data fusion: Supports over 50 data sources—from infrastructure and middleware to applications—to break down data silos.

• Visual dashboards: Customize views to show resource topology, service dependencies, and key performance indicators (KPIs).

• Cross-domain association analysis: Automatically link abnormal metrics with related logs and traces to identify root causes faster.

Scenario 2: Intelligent anomaly detection and failure prediction

During traffic spikes, manually identifying hidden failures is difficult. Cloud Monitor 2.0 uses machine learning to analyze historical data, predict risks like capacity bottlenecks and latency, and trigger early warnings.

• Root cause localization: Detect and compute in real time using metrics, traces, and profiling data. Covers latency, error rate, anomalies, and out-of-memory (OOM) cases.

• Impact scope analysis: Identify affected areas—such as end users, frontend applications, page requests, applications, interfaces, databases, and containers or ECS instances.

• Copilot self-service exploration: Use generative AI to retrieve detection reports and solutions on demand.

• Alert convergence: Converge alerts across products and instances. Group alerts with the same root cause to reduce noise.

Scenario 3: End-to-end full-stack tracing from client to server (APM)

In microservice architectures, a single request may involve dozens of service calls. Cloud Monitor 2.0 combines full-stack tracing with code-level diagnostics, linking user experience upstream to infrastructure downstream to precisely analyze slow queries and deadlocks.

• Full-stack observability graph: Covers services, interfaces, cloud service instances, and rich data—such as metrics, events, and metadata—with cross-domain entity associations.

• Associated data query and analysis: Upstream—retrieve real-time, dynamic upstream access endpoints and analyze business impact. Downstream—retrieve real-time, dynamic downstream dependencies—such as middleware, databases, and containers—with full monitoring data.

• Dynamic architecture awareness: Build a complete, global CMDB with auto-discovery and full topology views.

Scenario 4: Security compliance and threat insights

Enterprises need real-time monitoring of security events and compliance auditing. Cloud Monitor 2.0 uses log analysis and behavior pattern recognition to detect threats like abnormal logins and data breaches.

• Real-time threat detection: Use rule engines and AI models to identify attacks—such as abnormal logins and SQL injection.

• Compliance audit reports: Automatically generate operation log reports to meet compliance needs—such as China’s classified protection scheme and GDPR.

• Automated response: Integrate with security groups or WAF to block high-risk IP addresses automatically.

Scenario 5: Resource optimization and cost management

Opaque cloud resource usage leads to waste. Cloud Monitor 2.0 analyzes resource utilization and recommends elastic scaling policies and idle resource release plans.

• Utilization analysis: Flag low-load ECS instances and unmounted disks. Generate optimization checklists.

• Cost forecasting: Predict monthly bills based on historical spending trends and provide cost-saving tips.

• Automated elasticity: Scale Kubernetes clusters or Serverless services up or down based on traffic.

Scenario 6: Intelligent alerting and automated O&M

Traditional alerting often causes false positives or information overload. Cloud Monitor 2.0 uses alert noise reduction, dynamic thresholds, and tiered notifications to improve accuracy—and supports automated remediation actions.

• Alert aggregation: Merge similar events to avoid duplicate notifications.

• Multi-channel delivery: Push alerts by severity—such as to DingTalk, email, or text message.

• Automated scripts: Trigger prebuilt scripts to restart services or isolate faulty nodes.

Scenario 7: Managed open-source observability components and intelligent O&M

Enterprises widely use Prometheus, Grafana, and OpenTelemetry in hybrid or multicloud environments but face three challenges:

1. High O&M complexity: Self-managed Prometheus clusters require managing data collection, storage, and alerting end to end. Deployment and scaling are costly.

2. Data silos: OpenTelemetry trace data, Prometheus metrics, and Grafana dashboards reside in separate locations—lacking unified analysis.

3. Lack of intelligence: Open-source tools rely on manual alert rules and root cause analysis, struggling with dynamic AI-native architectures.

• Cost savings and efficiency gains: The managed service eliminates 90% of O&M work. Resource utilization improves by 30%.

• Full-stack observability: Covers infrastructure (Prometheus metrics), application performance (OpenTelemetry traces), and user experience (Grafana visualization).

• Open and compatible: Integrates seamlessly with open-source ecosystems—such as Prometheus Operator and Grafana plugins—to fit enterprise hybrid cloud stacks.

App type

App name

Description

Persistent

Alert Center

Manage all alert information in one place

Persistent

All Features

Manage all apps and related services in one place

Persistent

Integration Center

Integrate and manage observability objects and data

Resident

Entity Explorer

Explore the status and performance of monitored objects.

Persistent

Cloud Service Monitoring

Query and alert on basic monitoring metrics for Alibaba Cloud services

Application observability

Application Monitoring

Monitor application performance and diagnose faults in real time

Application observability

Real User Monitoring

Monitor web, mobile apps, and mini programs

Application observability

AI Application Observability

Deliver full-stack, integrated observability for AI applications

O&M monitoring

Prometheus Service

A fully managed Prometheus cloud service for high-performance monitoring

O&M monitoring

Incident Response

Group alert events into incidents and manage them

O&M monitoring

Synthetic Monitoring

Simulate user requests to proactively monitor network quality, service availability, and user experience

O&M monitoring

Database Observability

Deliver one-stop observability for database services

O&M monitoring

Log Audit

Record and review operation logs

Cloud service insights

PAI Insights

Deliver full-stack, one-stop observability for Platform for AI (PAI)

Cloud service insights

Container Insights

Analyze the operational status of Kubernetes clusters in depth

Cloud service insights

ECS Insights

Advanced monitoring for Elastic Compute Service (ECS)

Intelligent exploration and analysis

UModel Explorer

Entity and UModel debugging tool

Intelligent exploration and analysis

Data Explorer

Explore and analyze monitoring metrics and data

Intelligent exploration and analysis

Event Hub

Manage all types of event information in one place

Intelligent exploration and analysis

Dashboard

Dashboard showing key metrics

Intelligent exploration and analysis

Log Explorer

Provide log data exploration and analysis services