Teradata
Architecture
Teradata Database System are 2 Types A) SMP(Symmetric Multi-Processing)
B)MPP(Massively Parallel Processing)
A) SMP(Symmetric Multi-Processing): Teradata Database System is called
SMP System, if it have single NODE that contains multiple CPU Sharing Memory
Pool.
Node Architecture:
Node:
1)
Node
is important building block of the Teradata Database System and it is
collection of Hardware and Software Components.
2)
A
server can also be called as a Node.
3)
Channel
Attached System and Network Attached Systems can be connect to a Node.
4)
Channel
Driver and Teradata Gateway are the applications which runs under the Operating
System as Processors.
5)
Remaining
Components runs under PDE.
6)
PE
and AMP are virtual Processors, where BYNET is the Communication Layer between
PE and AMP.
PDE(Parallel Database Extention):
It Runs Teradata Components Parallel(PE,AMP and ByNet)
TPA(Trusted Parallel Application):
1) A Database is Pure Parallel or Trusted Parallel
Application, if it runs Under PDE.
2) Teradata is a Database which running
under PDE, so we call Teradata is Trusted Parallel Database.
Note: For Giga Bytes of Data Processing with minimum operations,
This ARCHITECTURE is Recommended.
PE(Parsing Engine):
1)
PE
Takes SQL Request and Delivers Sql Response.
2)
It
manages individual Sessions (up to 120).
3)
Parsing
and Optimizing your SQL Requests.
4)
Dispatching
the optimized plan to AMPs.
5)
Sending
the answer set response back to requesting
Client.
AMP(Access Module Processor):
1)
AMP is the
heart of Teradata which does most of the operations for data storage and
retrieval.
2)
Each AMP is
responsible for managing a portion of MAIN DISK SPACE(Virtual Disks).
3)
This space is
not sharable by any other AMPs so we call this Architecture is Shared Nothing
Architecture.
4)
Each AMP
contain Database Management Sub System, it perform below operations.
i)
Performing
DDL
ii) Performing DML
iii) Performing Joins
iv) Performs Applying and Releasing locks
v) Performing Aggregate Operations
vi) Performing OLAP Operations
vii) Performing SORTING Operations
BYNET (Banyan Net Work):
1) Carries messages between the AMPs and PEs
2) Provides Point-to-Point and Broadcast
communications
3) Merging answer sets and sends back to the PE
and Making Teradata parallelism possible
No comments:
Post a Comment