A unified representation for observable data
UModel is the data modeling layer of CloudMonitor 2.0. It gives every signal type—metrics, logs, traces, events, and profiles—a common schema so they can be queried, correlated, and analyzed together without format conversion.
Core Set types:
EntitySet: Defines a collection of entities. It includes primary keys, properties, and status fields.
TelemetryDataSet: The base type for all observable data. It provides a
timefield that MetricSet, LogSet, TraceSet, and EventSet inherit.LogSet: A flexible definition for log data with minimal structural constraints.
MetricSet: Defines a collection of metrics. It supports multiple metric values and tag dimensions in a single set.
TraceSet: A definition for distributed tracing. It includes fields such as TraceID and SpanID.
EventSet: A definition for event data. It supports both alert events and system events.
ProfileSet: A definition for performance profiling data. It covers CPU, memory, and other resource usage metrics.
Layered progression principle:
Each layer builds on the previous, so you only define what changes at each level:
Field: The most basic field definition.
ActivityData: Observable data with a time property.
Entity&TelemetryDataSet: Abstracted entities and observable data built on ActivityData.
Link: Connects entities to other entities and to data, forming a complete association graph.
Core idea of entity association and data fusion
Signals are most useful in context. UModel builds data graphs and entity graphs by associating signals with each other and with the entities that produced them—so a spike in a metric and a burst of errors in a log are no longer two separate observations, but two views of the same incident.
Data interconnection:
Logs and metrics share an entity context, making it possible to correlate log anomalies with metric anomalies in a single query.
Traces and logs are joined on TraceID, fusing distributed trace spans with the detailed log lines they produced.
Events are linked to entities, giving alert events and system status a unified view.
Entity interconnection:
A service-to-database call relationship lets you trace how database latency propagates into service response times.
A container-to-host containment relationship enables fault propagation analysis at the resource layer.
An application-to-infrastructure dependency supports full-stack root cause analysis across all layers.
Neuron-like connections: fault propagation and root cause analysis
UModel models system topology as a graph, similar to how neurons form a network. When a fault occurs, the Entity Relationship Diagram (ERD) acts as a map—tracing how the failure propagates and identifying the source.
Fault propagation model:
Simulates fault propagation paths through the relationship graph defined by EntitySetLink.
Supports multi-level traversal to trace from an abnormal entity back to the root cause entity.
Combines historical data and machine learning to score propagation probability at each hop.
Root cause analysis capabilities:
Alert events automatically associate with affected entities to generate a minimum connected subgraph—scoping the blast radius immediately.
Time series analysis identifies the chronological order of faults, distinguishing cause from effect.
Change events such as CI/CD deployments and configuration updates are correlated with faults, quickly surfacing change-induced regressions.
The path from data to insight
UModel structures raw telemetry into insight through five progressive steps:
Data normalization: Standardizes data formats using Field definitions and Sets.
Entity modeling: Organizes scattered signals into entities using EntitySet.
Relational association: Connects data and entities using Link to build the association graph.
Graph-based analysis: Runs complex association analysis and pattern recognition over the graph.
Intelligent insights: Applies AI algorithms to automate anomaly detection and root cause analysis.