Primary Cluster | VantageCloud Lake - Primary Cluster - Teradata Vantage

Teradata® VantageCloud Lake

Deployment
VantageCloud
Edition
Lake
Product
Teradata Vantage
Published
January 2023
Language
English (United States)
Last Update
2024-02-17
dita:mapPath
phg1621910019905.ditamap
dita:ditavalPath
pny1626732985837.ditaval
dita:id
phg1621910019905

A primary cluster is required in all VantageCloud Lake environments. You can think of the primary cluster as an Analytics Database with similar functionality and architecture. It consists of BYNET-connected nodes with Access Module Processors (AMPs) which process data from parsing engines, and move and manage data from disk drives. Each AMP owns its assigned data and has its own instance of the Block File System.

Because the primary cluster supports persistent block storage associated with each AMP, you can move data off an EDW platform and place it on a VantageCloud Lake primary cluster.

The vision of the VantageCloud Lake architecture is to support a single primary cluster and one or more compute clusters, with most of the work running on the compute clusters, which can be organized by department or application.

Teradata recommends gradually moving large tables for analytics and long-running queries, and tables that are objects of extract, transform, and load (ETL) processes, to the Object File System (OFS). For optimal performance and cost, you should store data used in tactical or short requests on the primary cluster.

General Purpose of the Primary Cluster

Here are ways that the primary cluster operates in this multi-cluster environment:
  • Entry point for all requests that enter a VantageCloud Lake environment.
  • Contains the data dictionary that describes all the tables, views, macros, functions, along with other software and VantageCloud Lake specific components.
  • The recommended location to run very short tactical queries when minimal latency is required.
  • Every query undergoes user security checks and validations, parsing, and optimizing on one of the primary cluster parsing engines.
  • All query locking steps and DDL steps such as CREATE TABLE are executed on the primary cluster.
  • The final spool file is built on the Primary Index before being returned to the client.
  • Landing location for TPT (Teradata Parallel Transporter) jobs loading updates into staging tables before updating tables in OFS storage.
  • Runs queries when the user does not have privileges to use a compute group and/or the data accessed by the query is stored on the primary cluster block storage.
  • A back-up location for queries to run when expected compute clusters are not available at the time the query runs.