98264270 Teradata Architecture

download 98264270 Teradata Architecture

of 27

Transcript of 98264270 Teradata Architecture

  • 8/12/2019 98264270 Teradata Architecture

    1/27

    Teradata ArchitectureTeradata Architecture

    AUTHOR: JAYAKRISHNAN.V

    E-MAIL : jayakrishnan.vijayakuar!"i#r$.%$

  • 8/12/2019 98264270 Teradata Architecture

    2/27

    Teradata corporationTeradata corporation

    Teradata corporation is a vendor specializing in datawarehousing and analytic applications .

    Its products are commonly used by companies tomanage data warehouses for analytical and businessintelligence purpose .

    Teradata was formerly a division of NCR Corporation

  • 8/12/2019 98264270 Teradata Architecture

    3/27

    FeaturesFeatures

    T&ra'a(ais Relational Database anagement !ystem

    "RD#!$

    %&sed for Data warehousing

    %'(ecutes on &ni( )*indows NT or *indows +,,,

    operating systems

    Teradata

    DATABASE

    )in *+)in NT

    IBM

    MainframeUNIX

  • 8/12/2019 98264270 Teradata Architecture

    4/27

    Compliant with -N!I standards

    Capable of running in single node or in multiplenodes

    &nlimited) roven !calability / amount of data and

    number of users0 allows for an enterprise widemodel of the data.

    &nlimited arallelism / arallel access) sorts) and

    aggregations.

    ature 1ptimizer / 2andles comple( 3ueries) up to45 6oins per 3uery) ad/hoc processing.

  • 8/12/2019 98264270 Teradata Architecture

    5/27

    odels the #usiness / 7N8) robust view processing) 9provides star schema capabilities

    :ow TC1 "Total Cost of 1wnership$ / ease of setup)maintenance) 9 administration0 no re/orgs) lowest dis; to dataratio) and robust e(pansion utility "recon fig$.

    2igh -vailability / no single point of failure.

    arallel :oad utilities / robust) parallel) and scalable loadutilities such as 8ast :oad) ulti:oad) and Tump.

  • 8/12/2019 98264270 Teradata Architecture

    6/27

    Parser Engine Parser Engine

    Message Passing Layer

    AMP AMP AMP AMP

    Vdisk VdiskVdiskVdisk

    Gateway S/

    Et!ernet Ada"ter

    TDP

    #LI

    #$ient %e&

    MTDP

    #LI

    #$ient %e&'

    M(SI#!anne$ Ada"ter

    B)s Ada"ter

    Netw*rk

    atta+!ed

    system

    #!anne$

    atta+!ed

    system

    Teradata Components 9 -rchitectureTeradata Components 9 -rchitecture

  • 8/12/2019 98264270 Teradata Architecture

    7/27

    Teradata RDBMS ComponentsTeradata RDBMS Components

    #y + ways we can connect to Teradata RD#!

    It creates sessions) allocates re3uests and response andalso fetches responses

    b$ TD % icro Teradata Director rogram

    > Controls the session related issues

  • 8/12/2019 98264270 Teradata Architecture

    8/27

    d$ 1!I % icro 1perating s?m interface

    /> It provides 1! independent interface

    8or connecting to Teradata RD#! networ; attached systemneeds 'thernet adapter

    ,hann& a((a%h&' sys(&

    C:I is having the same function as in n?w attached s?m TD % Teradata Director rogram

    anages session traffic b?n C:I n Teradata database) sessioninitiation and termination) logging) verification) session

    balancing etc 8or connecting to Teradata RD#! channel attached system

    needs host channel adapter and bus adapter in Teradata RD#!

  • 8/12/2019 98264270 Teradata Architecture

    9/27

    Teradata RDBMS componentsTeradata RDBMS components

    5 main components

  • 8/12/2019 98264270 Teradata Architecture

    10/27

    T!e Parsing Engine

    Answer Set

    %es"*nse

    Parsing

    Engine

    Message Passing Layer

    AMP AMP AMP AMP

    S,L %e&)est

    Parser

    ("timi-er

    Dis"at+!er

    The arsing 'ngine is

    responsible for@

    % anaging individual

    sessions "up to

  • 8/12/2019 98264270 Teradata Architecture

    11/27

    7 main components for '

    a/ars&r an' R&s$v&r

    Chec;s for synta( errors

    Chec;s the access permission for re3uested db ob6ect

    Chec; the e(istence of re3uested ob6ect and also returns error

    message if appropriate access is not there.

    0O#(ii1&r

    It prepares access plan and '(plain plan which shows how my3uery is solved efficiently means it restructure the 3uery in sucha way that it will run more efficiently

    %2&n&ra($r

    Ta;es the e(plain plan created by optimizer and converts it intodb understandable language called - steps

  • 8/12/2019 98264270 Teradata Architecture

    12/27

    +$ : % essage passing layer

    /> 2andles the internal communication ofTeradata dbms

    /> - steps are distributed to corresponding- based on the 2ashing algorithm over the

    essage assing :ayer "#AN'T$

  • 8/12/2019 98264270 Teradata Architecture

    13/27

    7$ - % -ccess module processor

    /> Birtual processors running under a multitas;ing

    environment

    /> #AN'T interface

    /> anage database

    /> Interface to dis; subsystem

  • 8/12/2019 98264270 Teradata Architecture

    14/27

    Th& AM/s ar&r&s#$nsi0& 3$r:

    - 4in'in5 (h& r$"s

    r&6u&s(&'- L$%k ana5&&n(- S$r(in5 r$"s- A55r&5a(in5 %$uns- J$in #r$%&ssin5

    - Ou(#u( %$nv&rsi$n an'3$ra((in5- ,r&a(in5 ans"&r s&( 3$r%i&n(- 7isk s#a%&

    ana5&&n(- A%%$un(in5- S#&%ia u(ii(y #r$($%$s- R&%$v&ry #r$%&ssin5

    T!e A++ess M*d)$e Pr*+ess*r

    .AMP

    AMPs st*re and retrie0e r*ws t* and

    fr*m disk

    Answer Set

    %es"*nse

    Message Passing Layer .PDE and B1NET

    AMP AMP AMP AMP

    S,L %e&)est

    Parsing

    Engine

  • 8/12/2019 98264270 Teradata Architecture

    15/27

    `5$ D' %arallel database e(tension

    This component is an interface layer on the top of

    operating system. Its functions include &8&%u(in5

    v#r$%s"virtual processors$) providing a #ara&&nvir$n&n() scheduling sessions) debugging) etc.

  • 8/12/2019 98264270 Teradata Architecture

    16/27

    Teradata St*rage Ar+!ite+t)re

    Teradata

    AMP 4AMP 3AMP 1 AMP 2

    Parsing

    Engine(s)

    Message Passing Layer

    18

    254

    41

    1290

    75

    80

    32 667

    25

    %e+*rds 2r*m #$ient .in rand*m se&)en+e

    2 32 67 12 90 6 54 75 18 25 80 41

  • 8/12/2019 98264270 Teradata Architecture

    17/27

    The arsing 'ngine dispatches re3uest to insert a row.

    The essage assing :ayer insures that a row gets tothe appropriate - "-ccess odule rocessor$.

    The - stores the row on its associated "logical$dis;.

    -n - manages a logical dis; which is mapped tomultiple physical dis;s in a dis; array.

  • 8/12/2019 98264270 Teradata Architecture

    18/27

    Teradata %etrie0a$ Ar+!ite+t)re

    Teradata

    AMP 4AMP 3AMP 1 AMP 2

    Parsing

    Engine(s)

    Message Passing Layer

    18

    254

    41

    1290

    75

    80

    32 667

    25

    %*ws retrie0ed fr*m ta3$e

    2 32 67 12 90 6 54 75 18 25 80 41

  • 8/12/2019 98264270 Teradata Architecture

    19/27

    The arsing 'nginedispatches a re3uest to retrieve

    one or more rows.

    The essage assing :ayerinsures that the

    appropriate -"s$ are activated.

    The -"s$locate and retrieve desired row"s$ inparallel access.

    essage assing :ayerreturns to retrieved rows to '.

    The 'returns rows to re3uesting client application.

  • 8/12/2019 98264270 Teradata Architecture

    20/27

    Teradata Parallelism

    'ach ' can handle up to

  • 8/12/2019 98264270 Teradata Architecture

    21/27

    M)$ti"$e Ta3$es *n M)$ti"$e AMPs

    EMPLOYEE ROWS

    EPAR!ME"! ROWS

    #O$ ROWS

    EMPLOYEE !a%&e EPAR!ME"! !a%&e #O$ !a%&e

    Parsing Engine

    AMP 4 AMP 5 AMP 6 AMP 7

    Message Passing Layer

    EMPLOYEE ROWS

    EPAR!ME"! ROWS#O$ ROWS

    EMPLOYEE ROWS

    EPAR!ME"! ROWS#O$ ROWS

    EMPLOYEE ROWS

    EPAR!ME"! ROWS#O$ ROWS

    !ome rows from each

    table may be found

    on each -.

    'ach - may have

    rows from all tables.

    Ideally) each -

    will hold roughly thesame amount of data.

    ultiple nodesM $ti N*de MPP S stem

  • 8/12/2019 98264270 Teradata Architecture

    22/27

    ultiple nodesmay beconfigured to

    provide aassively arallel

    rocessing "$system.

    - physicalmessage passinglayer called the

    #AN'T is neededto interconnectmultiple nodes.

    Teradata is alinearly

    e(pandableRD#! / as yourdatabase grows)additional nodesmay be added.

    M)$ti8N*de MPP System

    B1NET

    DA# DA#

    SMPSMPSMP SMP

    DA# DA# DA# DA# DA# DA#

    B1NET .f*r MPP

  • 8/12/2019 98264270 Teradata Architecture

    23/27

    BYNET Features:

    Ena!es mu!tip!e SM"nodes to

    communicate# Automatic !oad

    a!ancin$ o% messa$etra%%ic#

    Automaticrecon%i$uration a%ter

    %au!t detection# Fu!!& operationa! dua!

    BYNETs pro'ide %au!tto!erance#

    Sca!a!e and(idthas nodes are added#

    E'en thou$h thereare t(o ph&sica!BYNETs to pro'ideredundanc& andand(idth) Teradataand TC"*+" so%t(are

    on!& see a sin$!enet(or,#

    B1NET .f*r MPP

    T!e B1NET is a d)a$ red)ndant9 3i8dire+ti*na$

    inter+*nne+t netw*rk'

    A$$ SMPs are +*nne+ted t* 3*t! B1NETs'

    B1NET : B1NET 4

    SMP SMP SMP SMP SMP SMP SMP SMP

  • 8/12/2019 98264270 Teradata Architecture

    24/27

    T&ra'a(a 3i& sys(&T&ra'a(a 3i& sys(&

    +s a !a&er et(een Teradata RDBMS

    and "DE

    A!so pro'ides set o% ser'ice ca!!s thata!!o(s Teradata RDBMS to store and

    retrie'e data e%%icient!&

  • 8/12/2019 98264270 Teradata Architecture

    25/27

    7isk Arrays7isk Arrays

    - dis; array contains drive groups

    Drive groups contain set of drivers

    :&N ":ogical units$ contains a portion of every drive pdis; is a slice of :&N

    Froup of pdis;s assigned to an - is called v'isk

    7 ( 7i ( i0 (i7 ( 7i ( i0 (i

  • 8/12/2019 98264270 Teradata Architecture

    26/27

    7a(a 7is(ri0u(i$n7a(a 7is(ri0u(i$n -ccording to the primary inde( selected) the rows get

    distributed randomly among all the -s.

    ore the uni3ue the primary inde() the more even thedistribution will be.

  • 8/12/2019 98264270 Teradata Architecture

    27/27

    T-AN.S