«

»

Feb
24

Teradata Architecture


The biggest strength of the Teradata is the parallelism. So the architecture of the Teradata is designed in such way to keep this strength in mind. Teradata is unique from any other database because of its unique architecture only.

The main components of Teradata architecture is PE (Parsing Engine), AMP (Access Module Processor) and BYNET . We look into these components in details after looking the logical view of the architecture.

The logical view of Teradata architecture is given below -


Parsing Engine (PE) – Whenever a user login to Teradata it actually connect to Parsing Engine (PE). When a user submits a query, then the PE takes action, it creates a plan and instruct AMPs what to do in order to get the result from the query. The PE knows all, it knows how many AMPs are connected to Teradata system, how many rows are in the table and what is the best possible plan to execute the query. This is why the PE is also called as the ‘OPTIMIZER’.

Beside making a perfect plan for query execution PE also make a check on the access right of the user that weather the user has the privilege to execute the query or not. In this way PE also perform security feature on the users.


Access Module Processor (AMP) – Each AMP attached to the Teradata system listens to the PE via the BYNET for instructions. Each AMP is connected to its own disk and has the privilege to read or write the data to its disk. The AMP can be best considered as the computer processor with its own disk attached to it. Whenever it receives the instructions from the PE it fetches the data from its disk and sends it to back to PE through BYNET. Each AMP is allowed to read and write in its own disk ONLY. This is known as the ‘SHARED NOTHING ARCHITECTURE’. Teradata spreads the rows of the table evenly across all the AMPs, when PE asks for data all AMPs work simultaneously and read the records from its own DISK. Hence a query will be as slow as the slowest AMP in the system. This is known as parallelism.

BYNET – The BYNET is the communication channel between PE and AMP. It ensures that the communication between PE and AMP is correct and on right track. In Teradata system there are always two BYNET system. They are called as ‘BYNET 0’ and ‘BYNET 1’. But  we refer them as a single BYNET system. The reason two BYNET exist on a Teradata system is that –

1)      If one BYNET fails, the second BYNET takes over it place.

2)      Two BYNET improve the performance of the system, the PE and AMP can talk to each other over both BYNET which fasten the communication.


Short summary –

  • The PE checks the syntax of the query, check the user security rights
  • Then PE come up with the best optimized plan for the execution of the query
  • The PE passes this plan through BYNET to AMPs.
  • The AMPs follow the plan and retrieve the data from its DISK.
  • Then AMPs passes the data to PE through BYNET.
  • The PE then passes the data to the user.



I have also identified few E books which is immensely useful for Teradata Certification Preparation. Please have a look and if you are seriously preparing for TD certifications then don’t hesitate to purchase it.(Hover the mouse on the below image to get more information about the book)

 

 

81 comments

No ping yet

  1. Satish says:

    What is the maximum limit of AMP connected to BYNET?

    1. admin says:

      Each BYNET can support maximum 512 nodes.
      In Each Node we can run maximum up to 16384 VPROCS (AMP, PE, Gateways etc).

  2. Shruti says:

    Very nice explanation for each component. Should share some more infirmation about Teradata.

  3. admin says:

    Hi All,

    We have started a new special online teradata training batch for developers and DBA profile. For those who are interested to learn teradata can register in this batch. Fees is quite less when you compare it with other batches and biggest advantage is that its a instructor led training batch so you can ask any doubt in the training itself. Also we’ll be covering latest version of Teradata i.e. Teradata 14.10

    Please check following link for more details about this batch –
    http://www.onlineteradatatraining.com/?page_id=96

    Limited Seats. So try to register ASAP.

  4. Ramnath Jk says:

    If each PE supports up to 120 sessions and each session can run 16 requests in parallel and each AMP can run 80 tasks in parallel, does this mean that we need minimum 24 AMPs ((120*16)/80)???

    1. aditya says:

      Hi Admin,

      Could you please reply to this question i am also intrested in knowing the answer of this

      1. Lakshmi Srinivas Bodepu says:

        No of AMPS used depends on the type of request submitted and type of index used in the query. A request can be a single AMP,group AMP or an all AMP one based on whethere we are using a PI/SI or no index.

        No of AWT’s(AMP worker tasks) used depends on how many steps into which PE/Optimizer disects our query into.

        At most 62 AWT’s can run in an AMP to do user related works, Rest 18 are reserved for internal work types.

        At a single point of time a session can have only one active query(it can either be a single statement or a multi statement query). This single query is divided into multiple steps and sent to AMPS by PE, AMPS distribute these steps among AWT’s to do a portion of the work.

        For understanding an AMP can be compared to a Process and an AWT can be compared to a Thread running under a Process.

  5. shashank says:

    i need to learn basics of teradata as i am using this in mainframe… can please any body guide me which materials or url shd i use

    1. Shivam says:

      Shashank,

      Same situation here, Even I use this with MF. Please feel free to contact me – sriv.shivam@outlook.com
      I am using TD v12 with MF, in retail Project

  6. anilreddy says:

    this is best website ,good work guys.
    keep it .

  7. Arpan says:

    Awesome Explanation……………………………. Superr Guru

  8. Veena says:

    Thank you Sir! Your blog is very helpful, and the users too are intuitive because of which we get to learn new dimensions of a particular topic.
    I have one question-
    How is the number of AMPs , and size of each AMP decided?

    1. admin says:

      Number of AMP and size is decided at the time of configuration of teradata system. Usually it depends on the requirement of project.

  9. nilesh mali says:

    Nice Explanation for teradata Architecture

  10. Bharath says:

    very good explaination on TD for a novice.

  11. Indrajeet Patil says:

    Very informative…… Great work i must say..!!

  12. rajitha says:

    Hi admin.
    I wants to do teradata certification. please guide me

  13. UMESH says:

    WAT A SITE I NEVER SEEN BEFORE BEST EXPLANATION KEEP IT

  14. Jith says:

    Good site

  15. NEHA GUPTA says:

    hi

  16. hima bindu says:

    good explanation…

    I hav a doubt…

    according to u…Execution will be slow as much as slowest amp but when one amp fails what happens to execution…
    will it stop the entire execution..

    1. admin says:

      when an AMP fails then teradata data protection features comes in action like (hot standby node, cliques, down amp recovery journal).

      in case these protections are not able to migrate the data of that failed AMP (highly impossible) then your execution will stop until you recover that AMP.

      1. Sumit Nigam says:

        To further explain 2 points in case they are not clear:
        1. AMP writes to its “local” disk.
        2. In case, an AMP instance fails, a hot-standby can take over.

        Seeing both points together raises a question that if AMP writes locally then how does its local data get replicated to another instance. The way they do it is to actually have a common storage array between both AMP instances. They use iSCSI based SAN here, I think. So, both are sharing this “local” disk array. In case, the primary instance fails, the hot-standby can always be started without any “data-lag”. Of course, with iSCSI SANs also, storage tier can be a SPOF.

        1. Lakshmi Srinivas Bodepu says:

          It is correct that an AMP can perform I/O ops only on the VDISK/LUNS assigned to it.
          AMPS cannot read/write to VDISKS assigned to other AMPS.

          Coming to Hot Standby Nodes(HSN), They are protection mechanism for NODE failures, they can not protect against AMP failures.
          In case of a Node failure, AMPS get redistributed among other nodes or HSN(if present). For the migrated AMPS to perform operations on their VDISKS, Group of Teradata nodes(Cliques) are configured to access Disk Arrays of other nodes within the group like mentioned in previous comments. So, even if a node fails VPROCS migrated to other Nodes or HSN still can access the VDISKS in the Disk Array configured to the failed node.

          In case of AMP failures, The data of that particular AMP is not accessible which means full table scans and index operations involving failed AMP are not possible. Index accesses on healthy AMPS can still be performed.
          To protect against AMP failures, we have Fallback mechanism which replicates an AMPS data on another AMP. So, In case of an AMP failure, data can still be accessed from the fallback AMP.

    2. Vasu says:

      Hi Himabindu,
      Can you please share your mail id :vvdrmca@gmaail.com

  17. Gourav says:

    Hello Nitin,

    Thanks for putting up extra efforts to share your knowledge with us. keep it up.
    It is really good forum to learn TD concepts.

    Thanks,
    Gourav

  18. vivekanand kola says:

    Hi All,

    This is Vivekanand, this is very use full to who want learn teradata very quickly from basics to high level but it is very difficult to copy the data.Can any one give how to copy the data from this website.

    Thanks,
    Vivekanand kola

    1. admin says:

      Hi Vivekanand,

      Thanks for your comment.

      Actually copying content from this site is banned to maintain the originality of content.
      Hope you understand :)

      1. Mukesh says:

        Admin,

        We should not live in visual reallity . The basic concept for which Internet is created is sharing. I don’t understand when admin says that you can read from here but can’t copy from here .

        It’s just an CSS trick that disables right click which actually do not provide any security beacause any user who want data to copy will do ctrl + U and can copy easilly.

        request to admin this security is of no use and actually it annoys while typing becuase you can’t correct your mistakes by doing right click.

        Mukesh

  19. mujeeb says:

    Hello all..i m new in teradata today is my 1st day in this organization..this site helps me alot to increase my knowledge level about this organization..i will make sure that i will be helpful for the company and will help the company by using my skills.. ;)

  20. mujeeb says:

    Hey every one,i m new in teradata n this site helps me alot to increase my knowledge level about this organization..

  21. savitha subbaraman says:

    Hi admin,
    this site is really helpfull! can i know about AWT? and in the concept of index why we go for secondary index when we already have primary index?

    1. admin says:

      secondary index is the alternate way to access data. To know more about SI please refer – http://www.teradatatech.com/?p=815

  22. nitu says:

    Hi All,

    As I am new to TD and after reading the available input on this site really helped me to understand architecture of TD and how it actually works in real world in a very clear way to understand.

    Actually I am an SQL server resource but I am learning TD for informatica but found it more interesting so pls guide how to start with it in a more simpleway to understand it.

    1. admin says:

      if anyone is interested in teradata training for development profile please have a look – http://www.onlineteradatatraining.com/

  23. Anju Vishnu says:

    I find this site very helpful. All the concepts are presented in a simple manner.

    I have a doubt – Out of Node, PE, BYNET,operating system,CPU and disks , which is a part of teradata open architecture?

    1. priya says:

      Hi Anju,

      This question i have faced in TD12 basics exam.. NODE is the part of teradata open architecture.

      1. admin says:

        thats correct :)

  24. K.V.Koteswararao says:

    Hi,

    First of all Many thanks for your useful post. I am new to teradata DB. Can you please post about SMP and MMP architectures.

  25. Raju says:

    Hi ,

    The way you explained the architeture is really great and appricate.

    1. Ubendran says:

      Nice post for the beginners..

      1. admin says:

        Thanks :)

  26. Pavan says:

    Hello,

    Very nice post for the begineers..!!

    I write to get the right suggestion from you on my move to Teradata DBA role from Oracle DBA role, I have 6 yrs of experience as an oracle core DBA. I’m very much confused to take the decision. can you let me know the market and the oppurtunities in teradata for a DBA role. ‘ll the move benefit my career?

    Thanks
    Pavan

    1. admin says:

      Hi,

      Honestly speaking i don’t have much idea about ORACLE DBA market so can’t do the exact comparison between both these tech. But as far as TD DBA is concerned, there is huge requirements for DBA’s role, as you must be aware that TD is the emerging tech and most of the big shot clients are shifting towards huge data so its quite obvious that TD market is going strong day by day :)

      All the best for your career in TD

      Thanks

      1. KRISHNA says:

        GOOD ONE

  27. venu says:

    It’s very detailed and easily understandable for beginners

  28. Saurabh says:

    Thank You very much….
    I am very much waiting for more posts from you/

  29. ankita says:

    good for a beginner

    1. admin says:

      thanks .. will very soon post on advanced Teradata architecture.

  30. robert says:

    How do I find out how many AMPS my system has?

    1. rahul says:

      we can find the Amps in our system by using hash functions like
      1)hash Amp
      2)hash bucket
      3)hash row.. etc

      the query to find number of amps in our system is
      SELECT HASH AMP( ) + 1;

      here is a question that y we add 1. bcoz it shows one less than the actual amps. the reason is hash amp required the hash bucket value..

      Regards
      Rahul

      1. admin says:

        very well answer Rahul :)

        just to add one small point …. the actual syntax of HASHAMP is –
        HASHAMP (< hashbucket >)

        so it will give you the AMP number of the particular hashbuket value.

        but if you wont give any hashbuket value inside the bracket, then it will give the number of AMPs (less than 1) configured in the Teradata system.

  31. Pavab says:

    very good explanation
    thanks

  32. venkat says:

    how many amps will connect to bynet at at time?

    1. admin says:

      at the time all the AMPs in the teradata system are connected to BYNET.

  33. Chaitanya Datti says:

    very very clear explanation !!!!!!!!

    1. admin says:

      Thanks Chaitanya

  34. Suhasini says:

    Nice post.Thanks for sharing this

    1. admin says:

      Thanks Suhasini …

  35. Sunny says:

    Nice explanation, Admin. Appreciated !!

    1. admin says:

      thanks…. :)

  36. prakash says:

    good one for TD freshers….

    1. admin says:

      Thanks … :)

  37. Dilip says:

    Hi,
    Am new to tera data.can u send me the docs of tera data to my mail id.my mail id is dilip.thokala@gmail.com.

    Thanks & Regards
    Dilip.T

  38. Renuka says:

    Am new to teradata..This is very useful for me..Can you send the document for teradata? how to install & how to execute? Send this pdf of doc file to this mail id:renukadevi.ciet@gmail.com . Thanks in advance.

  39. PriyaSudheer says:

    We nice and crystal clear explanation….!!

    Dear Admin.. Topics mentioned are clearly expplained and very easy to quick grasp..please explain some more topics that are newly added to TD12 syllubus like (applaences,processing types etc,)
    Expecting positive reponse

    Thanks
    PriyaSudheer

    1. admin says:

      will definitely try to post on TD 12 :)

  40. sudha says:

    very clear explanation

    1. admin says:

      thanks .. :)

  41. Raki says:

    Really superb explanation!

    1. admin says:

      thanks :)

  42. vanitha says:

    nice explanation

  43. mw says:

    Neat. Clean, and I understood it. Thanks!

  44. prashant killada says:

    hai!
    That was superb contribution regarding the architectural summary of teradata.
    Thanks and regards
    Prashant killada

    1. admin says:

      Thanks Prashant for the appreciation :)

  45. Sadashivadas says:

    hi

    1. admin says:

      hi

      1. Ghouse says:

        Hi Admin,

        it is neat and clear explanation n understandable way.Thanks a lot :)

  46. admin says:

    hi Guys
    check for the new set of TD Basic Sample Questions available on my site –
    http://www.teradatatech.com/?p=325

    1. Vasu says:

      Hi Admin,
      I have one good question on to increasae Max number of sessions on TD 13.00
      Here my Environment is TD database 13.00 and OS :Win -2003..
      And We have 8 AMP’S and 1 PE on TD 13.00

      Now my question is
      We have a requirement of checking 200 simultaneous connection to Teradata.
      Currently we have only able to make max 120 connections. (Because 1 PE =120 Sessions)
      Now we need to increase the max no of sessions for Teradata DB. ?
      Can you please tell em how to increase max no of sessions on TD ..
      Is it possible to increase max no of sessions ?

      1. Lakshmi Srinivas Bodepu says:

        Vasu,

        It is not possible to raise the standard 120 session limit for a PE.

        Your objective of being able to connect 200 sessions is still achievable if you can add up one more PE to the system. This can only be done with the help of Teradata PS support team as this involves downtime and re-configuration of the entire system.

Leave a Reply

Your email address will not be published.

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=""> <strike> <strong>