Practice C14-037–048
C14-037 — Tools & Techniques · Foundational · Recall
What is the defining idea of an MPP shared-nothing architecture? - A. Partition data and computation across nodes with dedicated resources so work can run in parallel without shared disk/memory contention. - B. All nodes share one disk and one memory pool to maximize consistency. - C. Store every file on one server and use faster CPUs. - D. Use only streaming data and no historical data.
C14-038 — Tools & Techniques · Standard · Distinguish
A team needs inexpensive flexible landing for petabytes of structured, semi-structured, and unstructured files; immediate complex interrogative analytics is secondary. Which environment is the stronger fit? - A. MPP shared-nothing only. - B. Distributed file-based storage/database environment. - C. Traditional desktop spreadsheet. - D. Serving layer only.
C14-039 — Tools & Techniques · Standard · Recall
What are the three conceptual phases named for MapReduce? - A. Extract, Transform, Load. - B. Train, Validate, Test. - C. Map, Shuffle, Reduce. - D. Batch, Speed, Serving.
C14-040 — Tools & Techniques · Standard · Understand
Why can in-database algorithms be useful for Big Data analytics? - A. They eliminate the need for Data Quality and Metadata. - B. They guarantee a model will not over-fit. - C. They convert all unstructured data into relational data automatically. - D. They execute analysis close to the stored data, reducing unnecessary data movement and using the database platform's processing resources.
C14-041 — Tools & Techniques · Difficult · Distinguish
A workload is dominated by repeated high-performance parallel analytical queries over partitionable structured data. Which Chapter 14 tool family is most directly aligned? - A. MPP shared-nothing technology. - B. Distributed file landing only. - C. Data visualization tool. - D. Business glossary.
C14-042 — Tools & Techniques · Expert · Understand
A team rents a scalable platform temporarily to test whether a new external data source and model are economically worthwhile before committing to a permanent environment. Which Chapter 14 idea does this best illustrate? - A. Replacing business strategy with the cloud provider's feature list. - B. Using cloud/leased Big Data capacity to support exploration and feasibility before a larger landing/platform commitment. - C. Skipping governance because the environment is temporary. - D. Using the test set for iterative tuning because cloud experiments are non-production.
C14-043 — Governance / Implementation / Metrics · Standard · Understand
Which statement best reflects Chapter 14's principle about Metadata for Big Data? - A. Metadata is optional because Big Data schemas can be inferred when needed. - B. Metadata should be added only after a model reaches production. - C. Organizations should carefully manage Metadata about Big Data sources so they know what data exists, where it came from, and its value/context. - D. Metadata is only the physical file size and storage location.
C14-044 — Governance / Implementation / Metrics · Standard · Understand
Which is the best Chapter 14 reason that velocity does not justify skipping governance and quality controls? - A. Because governance always makes processing faster. - B. Because velocity is not one of the Big Data V's. - C. Because real-time data cannot be used for predictive models. - D. Fast data can still be wrong, misinterpreted, sensitive, biased, or poorly sourced; rapid processing increases the cost of propagating those failures.
C14-045 — Governance / Implementation / Metrics · Standard · Apply
Which governance decision is most directly implicated when analysts can freely combine internal customer data with external location data without review? - A. Data sourcing/sharing/access and security/privacy governance, including recombination risk. - B. Only query-performance tuning. - C. Only descriptive-vs-predictive classification. - D. Only model training-set size.
C14-046 — Governance / Implementation / Metrics · Difficult · Apply
A pilot model is accurate and useful, but the organization has no staff to operate the platform, no procurement path, and no business owner ready to change decisions based on the output. Which gate is failing? - A. Model-selection accuracy only. - B. Readiness assessment: organizational/business readiness, resourcing/talent, and operational/economic feasibility. - C. Data lake Metadata only. - D. Descriptive analytics maturity only.
C14-047 — Governance / Implementation / Metrics · Difficult · Distinguish
Which metric best demonstrates business value rather than only platform activity? - A. Number of files loaded per hour only. - B. CPU/storage hot spots only. - C. Revenue opportunity, cost reduction/avoided threat, useful model/pattern accuracy, or new business initiatives generated by the work. - D. Query count only.
C14-048 — Governance / Implementation / Metrics · Difficult · Apply
A deployed model's prediction accuracy remains stable, but the external source provider changes field meaning, the business process changes, and users stop acting on recommendations. What should monitoring cover? - A. Only prediction accuracy, because stable accuracy proves the model remains useful. - B. Only infrastructure uptime, because the model itself is unchanged. - C. Nothing; deployment completes the Data Science lifecycle. - D. Model performance plus source/Metadata changes, operational/business effectiveness, adoption/value, and the need to refine data, model, or process.