C1000-173 IBM Cloud Pak for Data V4.7 Architect Questions and Answers

To enable IBM Cloud Pak for Data to trust certificates from internal servers using a custom Certificate Authority (CA), the correct method is to create a Kubernetes ConfigMap that contains the CA certificate. This ConfigMap is referenced by the platform’s foundational services to include the CA in the trusted root store. Secrets are typically used for storing sensitive data like private keys and TLS certificates but are not used for adding trusted root CAs at the platform level. A ConfigMap is explicitly required by the platform to inject the CA trust into the certificate validation chain.

To integrate a supported vault system withIBM Cloud Pak for Data, thefully qualified URLof the external vault is a required component for establishing communication between Cloud Pak for Data and the vault. This is typically configured in theexternal secrets managersettings.

While authentication credentials (like client certificates or keys) are also necessarydepending on the authentication method used, thefully qualified URLis universally required to locate and connect to the vault.

IBM Cloud Pak for Data supports integration with vaults such as:

HashiCorp Vault

AWS Secrets Manager

Azure Key Vault

IBM Key Protect

For more details, refer to:

IBM Cloud Pak for Data: Using external secrets managers

In IBM Data Virtualization, caches must be manually defined by administrators. While monitoring and query performance statistics can guide where caching would be beneficial, the creation and configuration of caches (e.g., refresh schedules and scope) are manual tasks. There is no automated cache creation mechanism (A), nor are DataStage flows used for cache maintenance (B). Suggestions based on statistics (D) may assist administrators, but they do not automatically create the caches.

"You can create a maximum of10instances per deployment. After you reach the maximum number, the New instance button is not displayed in IBM Cloud Pak for Data."

A Data Refinery flow is an ordered sequence of data operations applied to tabular data for preparation, cleansing, and transformation. Users can create a series of steps such as filtering, joining, aggregating, and applying custom expressions. The flow is saved and can be rerun on updated datasets to ensure consistency in data preparation. It is not a storage system (A), an ML model (B), or a visualization tool (D).

After importing IBMKnowledge Accelerator(KA) assets using an API (or other methods), those assets — such as categories, terms, and relationships — are part ofgovernance artifactsin Cloud Pak for Data. However,to make them usable by specific users, youmust assign collaboratorsto the relevantcategories.

This ensures users have the appropriatepermissions(e.g., to view, curate, or manage terms) within the Information Governance Catalog (IGC) or Watson Knowledge Catalog.

The Data Engineer role in IBM Cloud Pak for Data includes permissions necessary for managing the data pipeline lifecycle. This includes the ability to create projects, manage data assets, and configure data protection rules, which are critical for ensuring data privacy and governance. The role does not include platform-level administrative privileges like managing the platform or monitoring all workloads. Workflow management is typically assigned to users with broader project or orchestrator roles.

Access to Match 360 within IBM Cloud Pak for Data is role-based and governed by service groups. Even administrative users are not granted automatic access unless they are explicitly assigned to the appropriate Match 360 service group. This allows fine-grained control over who can access master data management capabilities. Service group membership defines the roles and privileges needed for interacting with Match 360 functionalities like entity resolution and golden record management.

IBM Knowledge Catalog is the core governance and cataloging service within Cloud Pak for Data that enables the delivery of trusted, business-ready data to AI and analytics pipelines. It provides data lineage, metadata management, access policies, and quality scores, ensuring data consumers use curated and compliant data. Watson Machine Learning and its accelerator are focused on model training and inference, while IBM Data Catalog is a former term replaced by Knowledge Catalog in recent versions.

In air-gapped (offline) environments, it’s not just the container images that require preparation. Package managers like pip (for Python) and other utilities commonly attempt to pull dependencies from public internet repositories. These must be redirected to internal mirrors or handled via offline bundles. If left unaddressed, certain operations within components like Jupyter notebooks, Watson Studio, or pipelines may fail. License validation and monitoring do not inherently require internet access, and data movement between nodes is part of normal cluster function—not something to be blocked.

MPP, or Massively Parallel Processing, is a database architecture model where data is divided and processed across multiple compute nodes in parallel. Each node works independently on a portion of the data, dramatically improving query performance and throughput for analytics workloads. This model is ideal for big data and analytical queries, not transactional workloads. It differs from shared-disk models or replication strategies like two-phase commit. The correct definition involves distributed data and parallel query execution, as described in option C.

When selecting storage for Cloud Pak for Data, two critical considerations are:

Ensuring the storage supports the specific services being deployed. Not all services are compatible with every storage type (e.g., some services require block storage, others may support NFS).

The storage must provide sufficient I/O performance to meet the operational demands of the workloads.

Transmission speeds and throughput metrics are useful but not directly required as standalone criteria. Additionally, NFS is not universally supported by all services; some services specifically require RWO (ReadWriteOnce) or RWX (ReadWriteMany) access modes.

Watson Pipelines in Cloud Pak for Data support orchestration of diverse workload types including notebooks (Python or similar interactive environments) and scripts such as Bash. These pipeline components allow integration of notebook cells or shell scripts as tasks. There is no built‑in support for executing PowerShell tasks directly (unless wrapped in Bash-like containers), and Postgres is used as a data source—not a pipeline component type. While Db2 Big SQL can be invoked within a notebook or script, it is not itself a pipeline component. Therefore the supported types in pipelines are notebooks and Bash scripts.

Training Data Statistics:Watson OpenScale needs to understand the characteristics of the data the model was trained on. This includes things like the distribution of features, sensitive attributes (for fairness monitoring), and how the model performed on this initial data. These "training data statistics" are crucial for:

Fairness Configuration:Recommending fairness attributes, reference, and monitored groups.

Bias Detection:Calculating fairness metrics (like disparate impact) by comparing runtime behavior to the learned training data distribution.

Explainability:Generating explanations by understanding the distribution of values in the training data to create meaningful perturbations.

Drift Detection:Building a drift detection model that compares runtime data to the training data to identify shifts.

While Watson OpenScale also consumespayload data(the data sent to the deployed model for predictions) at runtime to calculate various metrics and perform monitoring, the initial setup and the ability to generate meaningfulstatisticsfor things like fairness and drift fundamentally rely on understanding thetraining data

InIBM Cloud Pak for Data version 4.7, when configuring Dynamic Workload Management (DWM) for IBM DataStage, the system controls job concurrency based on the maximum concurrent jobs setting and auto-scaling configuration.

Withauto-scaling disabled, the system does not add or remove DataStage engine pods dynamically to handle workload changes.

Themaximum concurrent jobssetting limits the number of jobs that can run simultaneously on a single DataStage instance.

If the number of concurrent jobs reaches the maximum limit (in this case, 5), any additional job requests (such as the sixth job) willnot fail immediately; instead, these jobs are placed in aqueue.

The queued jobs remain pending until one of the running jobs completes, freeing up capacity for the next job to start.

This queuing behavior ensures workload stability and prevents resource exhaustion by enforcing the concurrency limit strictly when auto-scaling is turned off.

Exact extract from IBM Cloud Pak for Data 4.7 documentation:

"When auto-scaling is disabled, the maximum concurrency limit set on the DataStage instance controls how many jobs can run simultaneously. Jobs submitted beyond this limit are queued and wait for running jobs to complete before starting execution."

—IBM Cloud Pak for Data v4.7, DataStage Dynamic Workload Management section