IBM Cloud Pak for Data v4.x Solution Architect Advanced Practice Exam: Hard Questions 2025

An enterprise is deploying IBM Cloud Pak for Data on OpenShift in two data centers (active/active at the OpenShift layer). They want uninterrupted access to the CPD web UI and APIs during a single data center failure, but they also require that user sessions and authentication remain consistent without forcing users to re-login. Which architecture best satisfies this requirement with the fewest CPD-specific risks?

A CPD cluster hosts multiple services (Watson Studio, WML, DataStage, Db2 Warehouse). Nightly batch scoring causes intermittent API timeouts and node pressure. The platform team wants to prevent model scoring workloads from starving governance and UI components, without overprovisioning the entire cluster. Which design is the most appropriate?

A bank must implement disaster recovery for CPD with a strict requirement: after a site loss, the restored environment must preserve governed data access policies and catalog assets so that data consumers see the same entitlements. They can accept some lag in analytics project artifacts. Which approach best aligns with CPD architecture and governance requirements?

A data engineering team uses DataStage on CPD to ingest from an on-prem RDBMS into a governed lakehouse. They must ensure that personally identifiable information (PII) is masked for most consumers, but a small group can see raw values for regulatory investigations. Which design best enforces this consistently across multiple tools (SQL queries, notebooks, BI) without duplicating datasets?

An organization uses CPD to orchestrate DataStage jobs that write to object storage and register outputs in a data catalog. They notice that some catalog assets appear without lineage, while lineage exists for others. The same job design is used, but failures are more frequent during peak hours. What is the most likely root cause and the best corrective action?

A team is integrating CPD with an enterprise messaging system for near-real-time ingestion. They need exactly-once processing semantics into curated tables, and they want replay capability for late-arriving corrections. Which solution pattern is the most appropriate on CPD, considering operational complexity and data correctness?

A regulated healthcare company trains models in Watson Machine Learning (WML) and must provide auditors with reproducibility: the ability to recreate a model exactly as deployed, including code, environment, and training data references. Which approach best meets this requirement?

A model deployed on WML shows intermittent latency spikes and occasional 5xx errors under load. The team suspects that autoscaling is not triggering quickly enough and that model initialization time is high due to large dependencies. What is the best corrective strategy that balances responsiveness and cost?

A data science team uses feature engineering in notebooks and wants consistent features between training and online inference. They observe training-serving skew because features computed in notebooks differ from those computed in the scoring service. Which architectural change most directly addresses the issue?

After integrating CPD with an enterprise LDAP/AD and enabling group-based access control, users can authenticate but intermittently lose access to catalog assets; re-login sometimes fixes it. Audit logs show policy evaluation failures during peak times. Which is the most likely cause and the best remediation?