Cloudera CDP Generalist key terms

Cloudera's integrated platform combining on-premises and cloud data services. It merges the former Hortonworks and Cloudera distributions under a single subscription, offering compute, storage, governance, and security components across multiple deployment form factors.

The on-premises deployment of CDP that packages Cloudera Runtime with Cloudera Manager for installation and management. It replaces legacy CDH and HDP clusters and runs on customer-managed hardware.

The cloud-native deployment of CDP delivered as managed services on AWS, Azure, and Google Cloud. Environments and Data Hub clusters are provisioned through the Cloudera Control Plane rather than by hand-installed software.

The collection of open-source components (such as HDFS, YARN, Hive, Spark, and HBase) packaged and versioned together as the distributed processing layer of CDP. Each CDP release pins a specific Runtime version.

The administrative application that installs, configures, monitors, and manages services and hosts in CDP Private Cloud Base. It provides a central UI and API for cluster operations and health monitoring.

A binary distribution format used by Cloudera Manager to install and upgrade software on cluster hosts. Parcels are versioned artifacts that are distributed, activated, and can be rolled back more safely than packages.

The Cloudera-managed service layer in CDP Public Cloud that provisions environments, manages credentials, and orchestrates Data Hub and Data Warehouse clusters. It runs outside the customer's cloud account.

A CDP Public Cloud construct that groups a cloud provider region, network configuration, storage locations, and a Data Lake under one credential and security context. Data Hub clusters are created inside an environment.

The shared foundational cluster in a CDP environment that hosts storage (HDFS or object storage), the metastore, governance services, and security components. Workload clusters depend on it for shared data and metadata.

A workload cluster in CDP Public Cloud that runs analytic services such as Spark, Hive, and Impala against the environment's Data Lake. Multiple hubs can share one Data Lake with isolated compute.

The CDP framework providing centralized security, governance, and metadata across clusters. It includes Ranger for authorization, Atlas for lineage and metadata, and the SDX namespace for shared context.

The authorization component in CDP that enforces fine-grained access policies on services such as HDFS, Hive, HBase, and Kafka. Policies are defined centrally and evaluated at query or access time.

A rule in Ranger granting or denying specific users or groups access to a resource, optionally with row filters or column masking. Policies are evaluated by service plugins before data is returned.

A Ranger policy feature that restricts which rows a user can see by attaching a SQL predicate to a table policy. It enables multi-tenant access to a single table without duplicating data.

A Ranger capability that transforms or hides specific column values for certain users, for example showing only the last four digits of an identifier. It is applied at query time rather than by altering stored data.

The metadata and governance service in CDP that captures technical metadata, classifications, and lineage for datasets. It underpins discovery and audit of how data flows between systems.

The recorded relationship showing how data moves and transforms from source datasets to derived outputs. In CDP it is captured by Atlas and used for impact analysis and audit.

A tag applied to Atlas metadata, such as marking a column as containing personal data. Classifications can propagate through lineage and be referenced by Ranger policies for dynamic protection.

The network authentication protocol used by CDP clusters to verify the identity of users and services through tickets issued by a key distribution center. It prevents impersonation on the cluster.

A file holding Kerberos principals and their encrypted keys, allowing services or scheduled jobs to authenticate without an interactive password. Keytabs must be protected because they grant the identity they contain.

An HDFS feature that encrypts data at rest within specified directories using encryption zones and key management. Clients read and write normally while data blocks and temporary files remain encrypted on disk.

An HDFS directory whose contents are transparently encrypted with a specific key from the key management server. Files created inside inherit the zone's encryption policy.

The Hadoop Distributed File System stores data across cluster nodes as replicated blocks managed by a NameNode. It provides fault tolerance and high-throughput access for batch processing.

The HDFS master service that tracks the file system namespace, block locations, and metadata. It does not serve file data itself and is a critical high-availability component.

An HDFS worker process that stores actual data blocks on local disks and serves read and write requests. DataNodes report block status to the NameNode and replicate blocks as instructed.

The number of copies of each HDFS block stored across DataNodes, commonly three by default. Higher values improve durability and read parallelism at the cost of storage.

The resource management layer that allocates CPU and memory across cluster nodes and schedules applications such as Spark and Hive. It separates resource management from the processing frameworks that use it.

The YARN master daemon that arbitrates cluster resources among applications and tracks node capacity. It works with per-node NodeManagers to launch and monitor containers.

The YARN agent running on each worker node that launches containers, monitors their resource usage, and reports to the ResourceManager. It enforces memory and CPU limits for tasks.

The unit of resource allocation in YARN, granting a bounded amount of memory and CPU to a task on a node. Applications request containers from the ResourceManager and run inside them.

A SQL-on-Hadoop service that maps files to tables and executes queries through an execution engine. It provides the metastore and a familiar SQL interface over distributed storage.

The service storing table definitions, partitions, and schema metadata used by Hive and other engines. Many CDP components share one metastore so tables are visible across tools.

A massively parallel SQL query engine that runs natively on Hadoop data for low-latency interactive analysis. It reads shared tables through the Hive metastore without moving data.

A distributed processing engine for batch and streaming workloads that keeps intermediate data in memory. In CDP it runs on YARN and is used for ETL, machine learning, and SQL via Spark SQL.

The Spark API for processing continuous data streams as unbounded tables with exactly-once semantics. It supports sources such as Kafka and sinks such as files or tables.

A distributed publish-subscribe log for streaming data, organized into topics and partitions with replicated brokers. It decouples producers from consumers and buffers high-volume event streams.

A shard of a Kafka topic that provides ordering within the partition and enables parallel consumption. Partitions are replicated across brokers for durability.

A set of Kafka consumers that cooperatively divide the partitions of subscribed topics. Each partition is assigned to one consumer in the group, enabling scalable parallel processing.

A scalable distributed object store in CDP designed for large datasets and small files, offering S3-like semantics on Hadoop. It can complement or replace HDFS for certain workloads.