Container Observability Capabilities

Stage 1 – Automated Discovery and Contextual Integration: Container Observability’s platform unifies the telemetry data (metrics, logs, traces, flows, etc.) in a distributed object model representing the microservice application as a dynamic distributed system. This enables representing application services of different types, whether microservice, SaaS, serverless, or FaaS, as well as understanding services can be auto scaled and work with ephemeral application components.

Stage 2 – Application Topology and Dependencies: After telemetry is processed and contextually mapped, Container Observability’s platform builds the application topology and all dependencies using eBPF network flow data without the need to enable tracing or using proprietary agents. This enables creating different views (‘maps’) of the complete application structure at the application level such as (App Map), which displays the service-to-service interactions and golden signals of performance per service, the Kubernetes node level (Node Map), the VM and the process (Host Map) as well as at the trace level (Trace Map).

Stage 3 – Flow and Trace Analytics are the two approaches that Container Observability uses for discovering the interactions and dependencies between components of the environment using flow analytics when code instrumentation is not available or possible, and distributed tracing: when the application is instrumented for tracing.

Stage 4 – Behavioral Profiling and Anomaly Detection: Container Observability ML continuously learns the behavior of each component in the application estate. Each deployment, DaemonSet, database, and cloud service has a unique behavior that is learned using curated model templates and a machine learning algorithm ensemble. The behavior of the component is compared to the predicted behavior and when there is a deviation, an alert is created.

Stage 5 – Automated Causal Analysis: the automated RCA conducts an AI-based diagnostics process that builds on all relevant information to the type of problem. Employing a dynamic decision tree process that uses curated knowledge (‘SME in a box') for problems in application to infrastructure, it leverages information on the current problem, configuration, and events, as well as learned dependencies and explanations from the behavior model. By querying conditions relevant to the type of problem, it can isolate the causal domain(s), and surface the relevant information to Ops so that they focus on the application object(s) that are the likely cause of the problem.

Stage 6 – Recommended Action: Once Container Observability isolates the problem source after it runs its automated causal analysis, a remediation action can be initiated. The level of automation that can be achieved depends on the nature of the problems, the operating environment, and the organization processes in place. This is currently not implemented.

Advent of Microservices

Overview