Node Memory | Advanced SQL Engine 17.10 - 17.10 - Node Memory Recommendations - Advanced SQL Engine - Teradata Database

Teradata Vantage™ - Advanced SQL Engine Release Definition

Product
Advanced SQL Engine
Teradata Database
Release Number
17.10
Release Date
July 2021
Content Type
Release Notes
Publication ID
B035-1725-171K
Language
English (United States)

For best performance, Teradata recommends that each node has at least the minimum recommended RAM.

When upgrading to Release 17.xx, several factors can cause some systems, especially large ones, to require additional memory. Follow the general guidelines; however, because memory requirements are workload-dependent, the actual memory requirements for your system may differ.

General Guidelines

  • Teradata recommends a minimum of 4 GB memory per vproc to achieve the most value and performance from SQL Engine 17.xx. The absolute minimum requirement is 2 GB memory per vproc, including AMP, PE, TVS, and GTW.
  • Some SQL Engine features require 512 GB per node; for example, Teradata In-Memory Optimizations.
  • Additional memory consumed is based on the size of system, AMPs per node, AWT, and feature use.
The maximum amount of memory allowed per node may be increased between database releases. Always double-check what the maximum memory amount is for your platform.
This release can run on a system with the recommended minimum RAM, but depending on the system configuration and the SQL Engine features you use, performance may not be optimal. You should also factor in the following to determine the optimal memory configuration:
  • Workload
  • Memory-consuming features
  • Performance requirements
  • Cost of memory

Memory Consuming Features

  • 1 MB Data Block
  • 1 MB Perm and Response Rows
  • 1 MB Phase 2
  • 1 MB Response Buffer
  • 1 MB Spool Row
  • 128K Parser Tree Segments
  • 3D Geospatial
  • Algorithmic Compression and Block Level Compression
  • Array INSERT
  • Auto Stats Enhancements
  • AVRO DATASET
  • BSON and UBSON
  • Columnar Primary AMP/Primary Index
  • Cylinder Read
  • Data Stream Architecture
  • DBQL – Show Parameters
  • Expanded Table Header
  • Extended Object Naming
  • External Stored Procedures
  • Geospatial Data Type
  • Geospatial Indexing
  • Global and Persistent Data (GLOP)
  • In-Memory Enhancements
  • In-Memory Optimizations
  • Increased Join/Subquery Limits
  • Incremental Planning and Execution
  • Java Stored Procedures
  • Join Index, Hash-Join, Stored Procedures, and 128K Data Blocks
  • JSON Data Type
  • Large Cylinder with Cylinder Read
  • Larger than 1 MB Plan Cache
  • LOBs and UDFs
  • More Than 20 AMPs/Vprocs per Node (All Releases)
  • More than 80 AWTs per AMP
  • Multiple Count Distinct Performance
  • Native Object Store
  • Online Archive Memory Enhancements
  • Parameterized Query Logging
  • Partial Online Reconfiguration
  • PPI and Multivalue Compression
  • Queryable Column Information on Views
  • QueryGrid: Teradata Database-to-Hadoop
  • Script Table Operators
  • Scripting and Language Support
  • SLES 11
  • SLES 12
  • SQL Interface for Ferret SHOWBLOCKS
  • Table Functions
  • Temporal DBS Support
  • Teradata Columnar
  • Teradata Intelligent Memory
  • Teradata Virtual Storage
  • Teradata XML
  • Tunable UDF Memory Limit
  • XML DBQL Logging
  • XSLT_SHRED_BATCH