Research & Evidence-Based Practice Flashcards

A cloud team is evaluating whether multi-cloud strategy reduces vendor lock-in risk. Which type of evidence would most strongly support or refute this hypothesis?

Answer: Longitudinal case studies documenting actual migration timelines, costs, and operational complexity from organizations that pursued multi-cloud

Longitudinal case studies from organizations that actually executed multi-cloud strategies provide real-world evidence of costs, benefits, and trade-offs over time.

When designing a load test to generate evidence for a cloud capacity planning decision, which factor is most critical to control?

Answer: Ensuring the load profile (request types, concurrency, think time) accurately reflects production traffic patterns

A load test only yields valid capacity planning data if the synthetic traffic accurately mimics real-world traffic distribution and patterns.

Which practice best embodies the 'measure twice, cut once' principle in cloud infrastructure changes?

Answer: Collecting baseline metrics before any change and defining rollback criteria, then deploying and comparing against baseline

Establishing a pre-change baseline and clear rollback criteria creates the evidence needed to evaluate whether the change achieved its goal without harm.

A cloud engineer wants to apply findings from Google's Site Reliability Engineering (SRE) book to their small startup. What critical thinking step is most important?

Answer: Assess which SRE principles apply at your scale and organizational maturity, adapting rather than directly copying Google's context

Practices from hyperscalers often embed assumptions about scale, team size, and resources that don't transfer directly to smaller organizations, requiring contextual adaptation.

What is 'Hawthorne effect' and why is it relevant to cloud performance experiments?

Answer: A bias where systems perform better during observation periods due to increased attention, potentially inflating experiment results

The Hawthorne effect can cause systems or teams to perform atypically during a monitored experiment period, making results less representative of normal operation.

A cloud team wants to build an evidence-based culture around incident response. Which leading indicator is best supported by research on high-reliability organizations?

Answer: Rate of near-miss reporting and action item closure from post-mortems

High-reliability organization research shows that near-miss reporting rates and post-mortem action item closure are the leading indicators most predictive of improving safety and resilience.

When publishing an internal research report on a cloud cost optimization experiment, which element most strengthens the report's credibility and reusability?

Answer: A methodology section detailing data sources, collection period, tools used, exclusions, and assumptions

A detailed methodology section allows others to critique the approach, reproduce the analysis, and understand the conditions under which the findings apply.