Skip to main content

Deployment Modes

Causely is designed to deploy directly into customer environments with flexibility to accommodate different security policies, observability stacks, and operational constraints. For most customers, Full Coverage Mode is the recommended deployment option. It is the easiest to set up, delivers comprehensive visibility, and produces the highest-confidence reliability insights.

Alternative deployment modes exist to support environments where host-level agents or streaming pipelines are not feasible. Each cluster or environment may use a different deployment mode if required.

How Deployment Modes Differ​

Deployment modes vary along three dimensions:

  • Coverage: how much system behavior is observable, including uninstrumented services
  • Freshness: whether insights are derived from streaming or periodically fetched data
  • Operational impact: agents, pipeline configuration, and ongoing operational ownership

All deployment modes preserve Causely’s core principles​

  • Deterministic, explainable causal modeling
  • Local processing whenever possible
  • Reuse of existing observability investments

Full Coverage Mode​

(Agent + Mediator)

Recommended for most customers.

Overview​

In Full Coverage Mode, lightweight Causely agents run on compute hosts and stream telemetry to a locally deployed Causely mediator. This allows Causely to observe system behavior directly at the host level, independent of application-level instrumentation.

What this enables​

  • Visibility into all services running on the host, including uninstrumented services
  • Reliable capture of short-lived and ephemeral workloads
  • Derivation of core service metrics (request rate, error rate, latency) directly from observed traffic
  • Ability to combine host-level observation with additional trace instrumentation (for example, OpenTelemetry) for runtimes or protocols that require deeper application context
  • Ability to enrich causal analysis with metrics fetched from existing metric backends (for example, Prometheus) when additional signal is required
  • The highest-fidelity causal model of system behavior with minimal blind spots

Why customers choose this mode​

  • Easiest path to meaningful results without auditing or modifying existing observability pipelines
  • Strong protection against gaps caused by missing instrumentation or aggressive sampling
  • Ideal for complex, fast-changing, revenue-critical environments

Trade-offs to consider​

  • Requires deploying an agent on each virtual machine or host
  • Agent lifecycle management introduces some operational overhead
  • Host-level access may require elevated permissions and security review
  • Non-VM workloads (for example, managed runtimes or serverless) may require additional integrations

Summary​

Full Coverage Mode provides a comprehensive and reliable understanding of system behavior and is the default choice for teams that depend on Causely to proactively assure reliability.

Streaming Integration Mode​

(Mediator Only, Push)

Intended for environments where host-level agents cannot be deployed.

Overview​

In Streaming Integration Mode, the Causely mediator is deployed in the customer environment and receives logs and traces from existing telemetry pipelines such as OpenTelemetry or Datadog.

What this enables​

  • Streaming analysis of logs and traces as data is produced
  • Reuse of existing instrumentation and collection infrastructure
  • Ability to enrich causal analysis with metrics fetched from existing metric backends (for example, Prometheus) when additional signal is required

Why teams choose this mode​

  • Avoids host-level agents due to security or operational constraints
  • Aligns with observability-mature environments
  • Provides timely insights using existing observability data pipelines

Trade-offs to consider​

  • Visibility is limited to already instrumented services
  • Detection quality depends on sampling, schema consistency, and data completeness
  • Infrastructure-level metrics may require additional metrics backend integration

Supported integrations (examples)​

  • OpenTelemetry streaming (traces and metrics)
  • Datadog dual shipping (traces; metrics depend on configuration)
  • All fetch-based integrations described in Backend Connect Mode

Summary​

Streaming Integration Mode enables real-time analysis using existing telemetry pipelines while trading off visibility into uninstrumented services.

Backend Connect Mode​

(Fetch from Internal Stores)

Intended for environments with a mature observability stack where existing telemetry pipelines should remain unchanged.

Overview​

In Backend Connect Mode, the Causely mediator connects directly to log, trace, and metrics backends and periodically fetches telemetry data for analysis.

What this enables​

  • Full reuse of existing observability pipelines and backends
  • No changes to agents, collectors, or exporters
  • Minimal operational disruption in tightly controlled environments

Why teams choose this mode​

  • Established, high-quality observability data already exists
  • Organizational policies restrict changes to production pipelines
  • Preference to leverage existing telemetry stores rather than introduce new data paths

Trade-offs to consider​

  • Telemetry is fetched on a schedule rather than streamed
  • Detection latency depends on polling intervals
  • Core service metrics must be mapped from existing metric stores
  • Queries may add load to customer-managed observability infrastructure
  • Effectiveness depends on sampling and retention policies in the underlying systems

Supported backends (examples)​

  • Prometheus (metrics)
  • Loki (logs)
  • Elasticsearch (logs)
  • Instana and Dynatrace (topology and metrics)

Summary​

Backend Connect Mode enables Causely to leverage existing, mature observability pipelines with minimal operational disruption, while trading off streaming immediacy.