Feb 24

Components of Teradata

Teradata is made up of following components –

Processor Chip – The processor is the BRAIN of the Teradata system. It is responsible for all the processing done by the system. All task are done according to the direction of the processor.

Memory – The memory is known as the HAND of the Teradata system. Data is retrieved from the hard drives into memory, where processor manipulates, change or alter the data. Once changes are made in memory, the processor directs the information back to the hard drive for storage.

Hard Drives – This is known as the SPINE of the Teradata system. All the data of the Teradata system is stored in the hard drives. Size of hard drives reflects the size of the Teradata system.

Teradata has Linear Scalability

One of the most important asset of Teradata is that it has Linear Scalability. There is no limit on Teradata system. We can grow it to as many times as we want. Any time you want to double the speed of Teradata system, just double the numbers of AMPs and PE. This can be better explained with the help of an example –

Teradata takes every table in the system and spread evenly among different AMPs. Each Amp works on the portion of records which it holds.

Suppose a EMPLOYEE table has 8 different employee id’s. Now in a 2 AMP system each AMP will hold 4 rows in its DISK to accommodate total 8 rows.


At the time of data retrieval each AMP will work on its DISK and send 4 rows to PE for further processing. If we suppose, one AMP will take 1 microseconds (MS) to retrieve 1 rows, then the time taken to retrieve 4 rows is 4 MS. And as we know that AMPs work in parallel, so both the AMPs will retrieve all 8 records in 4 MS only (4 MS time for each AMP).

Now we double the AMP in our system, and we use total 4 AMP. As Teradata distribute the records evenly among all AMPs, so now each AMP will store 2 records of the table.


Now according to our time scale, the time taken by each AMP for retrieving 2 records is 2MS.

So all 4 AMPs, working parallel, will retrieve the 8 records in 2MS only. Which was previously 4MS for the 2 AMP system.

Hence we double our speed by doubling the number of AMPs in our system.

This is the power of parallelism in Teradata. It is also known as ‘DIVIDE and CONQUER’ theory, according to which we are dividing the work equally and getting the result faster. To achieve the desirable speed we can increase the number of AMPs accordingly.

Related PostTeradata Architecture


Skip to comment form

  1. Navi


    If I double the AMP, performance would double. But, if I double the data and also double the AMPs, what would be the performance then?

    1. admin

      In that case, performance will be constant.
      you will not see any performance degradation because of double data.

  2. Sneha Naik

    Hi Admin,

    Could you explain the data distribution in the case of multiple tables.

    If I had 3 tables , tables A B and C.

    Table A has 2 records
    B has 7 records
    C has 10 records.

    How does the data distribution take place and how is it decided ?? Which component in the architecture decides on the distribution of rows of data.


    1. admin

      Each AMP stores a certain portion of each table. Meaning, ideally each AMP will have some rows of each table which is defined in Teradata (exceptions possible).

      To know more about the data distribution please move to the topic – http://www.teradatatech.com/?p=470

      Online Teradata Training –

  3. aiman sarosh

    Hi Admin,
    Great article..really helpful. but i have a question here about AMP and PE.
    Does PE and AMP work like master-slave architecture?…like in hadoop (name-node and data-node) ?. if yes, then how can we increase the number of PE, since we can have only one master and multiple slaves ?

    I am a hadoop SME, so trying to correlate the two.


    1. admin

      in teradata AMP and PE both are actually virtual processors (VPROCS), number of VPROCS in Teradata can be modify with actual NODE processor and memory.


    I m new to tera data ,can u suggest me some good ebooks for teradata.

    1. admin

      you can go for the e-books by Tera-Tom or material available on Teradata website itself.

  5. Gokul

    Good !!!
    Increasing PE s or AMP s , does it reflect in any other parameter increase ?/

  6. Gokul

    Good !!!
    Increasing PE s or AMP s increases any cost?

  7. Vladek

    I hate unexplained acronyms. ‘AMP’, ‘PE’ ????

    1. admin

      AMP = Access module processor
      PE = Parsing Engine

  8. Shirish

    Great job Admin.

    I have a question here.
    you are saying we can double the AMPs and PEs ( while u are explaining about linear scalability )
    Can we also double the PEs ? Bcoz, I can’t see multiple PEs anywhere in pictures.

    Please explain.


    1. admin

      Hi Shirish,

      Yes we can double PEs as well because PE is also one of the VPROC(virtual processor) running inside the node memory.

  9. Deepanwita Sarkar

    Hi Admin,

    Its really a good article. Wish to learn more .


  10. nilesh mali

    Fantastic explanation…………..great job…………….

  11. Shuhbodeep

    It was really a great presentation. However can you explain how hashing algorithm works in TD?? I have been listening this HA word from past one year but never got a single positive response from any of the ADMIN. Hope this will be my last post for this question

    1. admin

      the hashing algorithm is the usual hashing process available in any programming language. This is implemented as a C code inside Teradata optimizer. If you want to know about the code a quick google on ‘hashing algorithm example’ will help you.

  12. Monisha

    Dear Admin,
    Even I would like to take the classes. Please let me know the procedure,

  13. Laxmi, M.S.

    I am an admn. person. But somehow I want to learn teradata as I was challenged by friends that I can’t learn Teradata without any relative background. But I feel that data does not relate or belong to any particular field and it is a general and at the same time special and very powerful tool of Data Management and its uses. Depending upon the grey matter of the person handling this tool, I strongly believe that “Sky is the only limit” for its usage. This is the basic reason why I want learn this “Wonderful and Evergreen” subject. I hope my ambition is appreciated in the right spirit.

    I am a novice to IT field in general and TeraData in particular. I am a B.Com. graduate and superannuated from Central Government service recently. I want to utilise some of my time now for learning this useful subject. Thanks.

    Laxmi, M.S.

    1. admin

      Hi Laxmi,

      Your amibition is great and yes you have said all correct words 🙂

      btw are you interested in taking Teradata classes ?
      let me know …


      1. praveen kumar reddy

        Hi Admin,

        I am intrested to attend the classes. please let me know the procedure.
        Thanks for your articals which are well benifited for beginners like me. 🙂


  14. rashid

    do you have a printable versions of he tutorials so that i can print them out and study them when ever i want and highlight the main ideas of the article

    1. admin

      Now you can take the print out directly from the website.

  15. Abhishek


    Am newbie on teradata and exploring it .

    With in the AMP does teradata has MPP and shared nothing functionality ?

    In other words for single AMP does teradata has different server process having their own memory and disks.

    From explanation it is cleared that PE doesn’t contain any user data i hope 🙂 but How HA is achieved in teradata,is it like data residing on one AMP the same data would be present on another AMP in case of node failure to achieve no single point of failure.

    The basics concept is mentioned very clearly and thanks for that but a bit explanation on my doubts would be appreciated.


  16. PriyaSudheer

    Fantastic explanation….!!! gr8 job….

  17. Plastic Surgeon

    Kudos for the great article. I am glad I have taken the time to learn this.

    1. admin

      thanks 🙂

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>