01
Design
Build systems that tolerate disruption, degrade deliberately, and recover predictably.
Failure boundaries · graceful degradation · resilience by designEngineering resilient software
Resilience is afeedback loop.
SRE Norns builds free, source-available tools and publishes practical ideas for designing resilient systems, measuring how they behave, and continuously improving what happens next.
01The resilience loop
Operations is a system of action, not a wall of graphs.
SRE is applied control theory: define an outcome, measure the system in operation, compare reality with intent, then act. A measurement becomes useful when it helps the system resist, recover, or learn from disruption.
REFERENCE / 01Desired outcomeService-level intent
CONTROLLER / 02Decide & actSRE · automation · AI agents
SYSTEM / 03Service / plantSoftware in its environment
DISTURBANCE↓Load · failure · change
MEASUREMENT / 04Observe behaviorTelemetry · tests · outcomes
OUTER LEARNING LOOPIncident → Postmortem → Improve the loopFind gaps in measurement, decisions, automation, and design.
02The practice
Design. Observe. Adapt.
Resilience is not a property added after launch. It is a repeating engineering discipline that connects architecture, operations, and learning.
02
Observe
Measure service behavior and user outcomes—not component availability alone.
Continuous verification · synthetic monitoring · service outcomes03
Adapt
Use evidence, automation, and experimentation to improve what the system does next.
Control actions · AI agents · postmortem learning03Tools for the loop
Observe reality.
Improve deliberately.
Two early-stage tools turn the SRE Norns ideas into working practice. Both are free for permitted noncommercial use, with publicly available source.
URTH
Test service behavior from the networks where the service actually runs.
Run versioned HTTP, TCP, DNS, ICMP, gRPC, REST, HAR, and browser scenarios on workers inside private or segmented networks.
- Outbound-only self-hosted workers
- Label-routed execution across network boundaries
- Results, logs, metrics, screenshots, and artifacts
EXP-BENCH
Coordinate measurable improvements across human and AI engineering agents.
Turn repeated experiments into a shared record of hypotheses, evidence, reviews, integrations, and outcomes.
- Bring your own agent and execution environment
- Leased work across focused research roles
- Evidence trails from proposal to adopted change
Public previewThe demos are early environments, not production services. Data may be reset; do not upload production data, credentials, or secrets.
04Ideas & field notes
The theory beneath the tools.
Practical writing on service behavior, feedback, failure, and the engineering habits that make systems easier to recover and improve.
01
Explore all documentation Foundations
SRE as Applied Control Theory
A practical model for desired outcomes, feedback, disturbances, and corrective action.
02Observability
What Do We Observe?
Why behavior and user outcomes matter more than a map of internal components.
03SRE practice
Fighting Availability Bias
The most available metrics are not necessarily the measurements that matter.
04Reliability
Your Health Check Is Not Enough
Healthy components do not prove that the user-facing service is working.
05Built in public
Follow the threads with us.
SRE-Norns is in its early days. Try the demos, inspect the source, challenge the ideas, and help shape tools for evidence-driven resilience engineering.
Interested in commercial use? Talk to us about licensing.