Data Warehouse - Bitmap Indexing

download Data Warehouse - Bitmap Indexing

of 24

Transcript of Data Warehouse - Bitmap Indexing

  • 7/25/2019 Data Warehouse - Bitmap Indexing

    1/24

    Data Warehousing andData Mining

    Data Warehouse Design

    Physical Design

    Bitmap Indexing

  • 7/25/2019 Data Warehouse - Bitmap Indexing

    2/24

    Contain a large number of dataorganised in a table structure, i.e. inthe form of rows and columns

  • 7/25/2019 Data Warehouse - Bitmap Indexing

    3/24

    Read Write Tables (High volume useof Insert, !date, Delete statements"

    Read #nl$ Tables % &tatic Data

    'ggregate Tables

    Dimension Tables

    act Tables

  • 7/25/2019 Data Warehouse - Bitmap Indexing

    4/24

    Inde)es hel! to *uic+l$ locate !ieces ofinformation about a to!ic which arescattered around a number of

    resources. or e.g, the inde) of a boo+tells what !ages of the boo+ refer to aconce!t listed in the inde)es.

    Inde)es are created in DatabasesDatawarehouses for e-cientidentication of the re*uested rows

  • 7/25/2019 Data Warehouse - Bitmap Indexing

    5/24

    the act of classif$ing and !roviding an inde)in order to ma+e items easier to retrieve

  • 7/25/2019 Data Warehouse - Bitmap Indexing

    6/24

    / Tree (/inar$ Tree" Inde)ing

    Hash Table

    /itma! Inde)ing

  • 7/25/2019 Data Warehouse - Bitmap Indexing

    7/24

    0ormall$ used for #1T2 a!!lications(man$ read3write o!erations"

  • 7/25/2019 Data Warehouse - Bitmap Indexing

    8/24

    /3trees are alwa$s height balanced, with all leafnodes at the same level

    !date and search o!erations a4ect onl$ a few dis+!ages, so !erformance is good.

    /3trees +ee! related records on the same dis+ !age,which ta+es advantage of localit$ of reference.

    /3trees guarantee that ever$ node in the tree will befull at least to a certain minimum !ercentage. This

    im!roves s!ace e-cienc$ while reducing the t$!icalnumber of dis+ fetches necessar$ during a search oru!date o!eration over man$ thousands of records.

  • 7/25/2019 Data Warehouse - Bitmap Indexing

    9/24

    Contains a 5fact table5, !lus a numberof 5dimension tables5 whose +e$s areforeign +e$s of the fact table.

  • 7/25/2019 Data Warehouse - Bitmap Indexing

    10/24

    How would $ou e-cientl$ answer the*uer$6 5ind the total sales ofraincoats for each location last $ear.5

    7ou ma$ assume suitable inde)es.

  • 7/25/2019 Data Warehouse - Bitmap Indexing

    11/24

    0umber of relations can be large %8oin si9e estimation is di-cult.

    Dimension tables are often small,es!eciall$ when considering onl$ thetu!les matching a select o!eration.

    #ften data can be assumed to be

    static (a sna!shot of the o!erationaldata". This means that we have theo!tion to !recom!ute.

  • 7/25/2019 Data Warehouse - Bitmap Indexing

    12/24

    &u!!ose there are onl$ : di4erentlocations in our !revious e)am!le.

    Then we ma$ re!resent the locations of 0

    fact table tu!les using onl$ ;0 bits. However, a (clustered" inde) on location

    seems to re*uire at least 0 log 0 bits.

  • 7/25/2019 Data Warehouse - Bitmap Indexing

    13/24

    or each !ossible value(of 5location5" and eachtu!le, store a bit that is