SRE Foundation study plan

Define reliability from an SRE perspective and distinguish it from availability, durability, and uptime in your own words

Map each core SRE practice (SLIs/SLOs, error budgets, toil reduction, automation, monitoring, release engineering, simplicity) to one concrete example from a system you have operated

Explain the 100%-reliability fallacy and compute why each additional nine raises cost faster than it raises user value

Compare SRE, DevOps, and traditional operations across at least four dimensions (ownership, metrics, change management, failure posture)

Write three SLIs for one service using the ratio form (good events / valid events) and state where each is measured

Convert each SLI into an SLO by adding a numeric target and a compliance window

Calculate the error budget in minutes and in event counts for a 99.9% target over both a 28-day and a 30-day window

Write an error budget policy that names the trigger conditions and required actions at budget healthy, 50% consumed, and exhausted

Distinguish SLI, SLO, SLA, and error budget and explain which one is contractual

Design instrumentation for one request path so it yields request rate, error rate, and a latency distribution rather than an average

Select p50/p90/p99 thresholds and justify each choice for a user-facing service versus a batch pipeline

Sketch a four-golden-signals dashboard for one service and specify the query and threshold behind each panel

Rewrite two cause-based alerts as symptom-based alerts and state the user-visible condition each detects

Explain how alerting on SLO burn rate reduces noise compared with static threshold alerting

Log every recurring operational task for one week and quantify hours per week, frequency, and manual step count

Compute your current toil percentage and compare it against the 50% ceiling

Rank at least three automation candidates by effort to build versus recurring hours saved per quarter

Automate one repetitive task end to end and measure time per execution before versus after

Classify your tooling gaps into configuration management, CI/CD, observability, and load/chaos testing

Write an idempotent, versioned runbook for a routine operational task that a peer can execute without asking questions

Design a canary release with explicit promotion criteria and an automatic rollback trigger

Configure or specify one health check and one automated remediation action with a safe abort path

Evaluate one build-versus-buy tooling decision against at least three weighted criteria

Identify where automation removes toil versus where it merely moves it, using one example from your own stack

Write a blameless postmortem for a past incident including timeline, contributing factors, and impact in user terms

Separate root cause from contributing factors and derive at least three systemic fixes rather than one individual fix

Design a chaos experiment with an explicit hypothesis, bounded blast radius, and an abort condition

Explain three resilience tactics (redundancy, graceful degradation, load shedding) with a concrete implementation each

Describe how SRE interfaces with product, development, and leadership and name two likely sources of friction

Compare embedded, centralized, and consulting SRE team models in a tradeoff table covering ownership, scaling, and cost

Draft an SLO review cadence with a standing agenda and named participants

Argue the business case for error budgets to a non-technical stakeholder in five sentences or fewer

Compare SRE with DevOps, ITIL, Lean, and Agile on goals, metrics, and primary artifacts

Summarize three emerging directions (e.g. platform engineering, AI-assisted operations, cloud cost-aware reliability) with one risk each

Take a full-length timed practice exam under exam conditions and score at least 80% overall

Re-drill every missed item until you can write the correct reasoning without looking at the answer

Cross-check your SLO and error budget notes against the exam blueprint and close any gaps in domains scoring below 70%

Compress all eight domains into a single-page reference you could redraw from memory