TERADATA is unique because of it's parallel architecture. So it is very important to understand the Teradata Architecture very well. Understanding of architecture may help to understand how SQL Queries once submitted are processed by TERADATA. The execution steps can be checked with the help of Explain plan. TERADATA Architecture can be very easily explained with the help of below diagram.

The above diagram shows a very basic archiecture of TERADATA and is very useful and easy for learning purpose. The first component is PE (Parsing Engine). It has 4 components in it :
1) Session Control –> It checks for user authorization before processing any SQL queries. A PE can support upto 120 sessions.
2) Parser –> It checks for the SQL syntax and user rights to access various database objects referred in the SQL query submitted by user.
3) Optimizer –> It create a plan or execution steps to follow in order to perform actions on database objects as per SQL query submitted by user.
4) Dispatcher –> It passes the execution steps to BYNET. Dispatcher also plays an important role by combining all the responses received and send it to user.
Next component is BYNET. It is used for communication between PE's and AMP's (Access Module Processor). There are two BYNET's available in any TERADATA environment: BYNET-0 & BYNET-1. Two BYNET's allows for continuous sending and receiving messages between PE's and AMP's.
Next to BYNET, we have AMP's. AMP's can be considered as the worker in TERADATA Architecture. Each AMP has it's own dedicated VDISK (Virtual DISK) to which it queries and process the steps planned by Optimizer. AMP's work only on their own VDISK and do not have access to other AMP's VDISK. Once the AMP perform the steps, it send back response to PE via BYNET where all the responses from various AMP's is collected and sent back to user.