DWHPro Insights

Understanding Teradata Nodes in Parallel Systems: Architecture and Scalability

By Roland Wenzlofsky ·

Understanding Teradata Nodes in Parallel Systems: Architecture and Scalability

 

What is a Teradata Node?

This describes a classic Teradata appliance arrangement. Nodes, cabinets, physical storage and network topology differ across appliance, virtualized and cloud deployments.

On each node, the primary processes of Teradata systems are performed (see our article about the Teradata high-level architecture):

Parsing engines, AMPs and BYNET communication services run within the deployment; physical network redundancy is platform dependent.

Parallelism within a node is achieved by uniformly distributing the workload among all AMPs.

Teradata architecture offers excellent scalability, allowing numerous nodes to connect into a vast system.

The idea of achieving a performance boost by doubling the number of nodes is often called linear scalability. However, this is largely a myth in practice. The flaw becomes apparent when you consider the requirement for perfect parallelism across your workload.

Teradata Query Performance Tuning: The DWHPro Guide book cover

Get the complete Teradata tuning book, free

Teradata Query Performance Tuning: The DWHPro Guide
By Roland Wenzlofsky & Artemios Vogiatzis

206 pages on statistics and demographics, access paths and indexes, query execution plans, and joins. Get the full PDF and practical Teradata updates from DWHPro.

Confirm your email to receive your download link. Privacy policy. Unsubscribe in any newsletter.

About the free book

In practice, workload skew is a well-known problem. Adding more nodes will not improve efficiency if a SQL statement’s workload remains on a single AMP. This is an important consideration when expanding your system.

Scalability and fault tolerance depend on platform limits, workload, fallback/storage protection and restart or clique-migration configuration. Shared-nothing architecture is not inherently non-resilient, and no universal node ceiling follows from this overview.

In parallel system terminology, a single node is referred to as a symmetric multiprocessing (SMP) node. A system comprised of at least two nodes is classified as a massively parallel processing (MPP) system.

The communication network, BYNET, is software-based within a single node but hardware-based between nodes, facilitating communication between AMPs and Parsing Engines across different nodes.

Redundant BYNET arrangements depend on the hardware and deployment; two physical BYNETs are not universal.

Network redundancy can maintain service across some failures, but availability depends on the actual topology and recovery mechanisms. Do not generalize a two-network failure rule across deployments.

Years ago, Teradata’s BYNET provided a notable advantage by sorting and merging data, thus reducing CPU workload. However, this benefit may no longer be as significant with the advent of multicore processors. The shift from BYNET to InfiniBand as the primary data transmission backbone is a contributing factor.

Deciding what to change in your data platform?

Compare optimisation, modernisation and migration against your workload and business requirements.

Explore independent platform assessments

Teradata Query Performance Tuning: The DWHPro Guide book cover

Get the complete Teradata tuning book, free

Teradata Query Performance Tuning: The DWHPro Guide
By Roland Wenzlofsky & Artemios Vogiatzis

206 pages on statistics and demographics, access paths and indexes, query execution plans, and joins. Get the full PDF and practical Teradata updates from DWHPro.

Confirm your email to receive your download link. Privacy policy. Unsubscribe in any newsletter.

About the free book

Follow DWHPro in Google to see our articles more often in Search.

2 thoughts on “Understanding Teradata Nodes in Parallel Systems: Architecture and Scalability”

  1. Hi Falcon. Thanks. I added the link.

    Regarding your question: Yes, each AMP has its own working memory. The so-called FSG-cache of each AMP is used to hold the data blocks read from disk.

    Reply
  2. Roland – Thanks for an informative article. I have 2 comments/questions:

    1. Can you also include the link in the article to ‘Teradata high-level architecture’ that you have referred to in your second paragraph?

    2. How is TD’s RAM organized? Does each AMP have its own memory in addition to dedicated disk space? If not, can TD still be considered a shared-nothing architecture RDBMS?

    Reply

Leave a Comment

Teradata Query Performance Tuning: The DWHPro Guide book cover

Get the complete Teradata tuning book, free

Teradata Query Performance Tuning: The DWHPro Guide
By Roland Wenzlofsky & Artemios Vogiatzis

206 pages on statistics and demographics, access paths and indexes, query execution plans, and joins. Get the full PDF and practical Teradata updates from DWHPro.

Confirm your email to receive your download link. Privacy policy. Unsubscribe in any newsletter.

About the free book

Protected by Security by CleanTalk and CleanTalk Anti-Spam