ORACLE olap - dimension with huge number of members.

Hi Experts,
I have modeling a ORACLE OLAP cube with 10 dimensions. I have one of the dimensions which have huge leaf members...(probably 1 billion).. How many dimension members i can have for one dimension. Is there any limitation to the dimension members ?
Thanks and Regards
Siva

Hi there,
I'm not sure of the theoretical limits, but I would never recommend trying to build a dimension with 1 billion members.
In situations were such granular data exists, it is common to leave this in table and load data into the OLAP cube at a more aggregate level.
Which business entity are you trying to model in a 1 billion member dimension?
Thanks,
Stuart Bunby
OLAP Blog: http://oracleOLAP.blogspot.com
OLAP Wiki: http://wiki.oracle.com/page/Oracle+OLAP+Option
OLAP on OTN: http://www.oracle.com/technology/products/bi/olap/index.html
DW on OTN : http://www.oracle.com/technology/products/bi/db/11g/index.html

Similar Messages

Does Oracle OLAP comes with Oracle Database 10g ?

1.Does Oracle OLAP comes with Oracle Database 10g or do we need seperate software to have this ?.
2.If I create a Cube with Analytic work space manager the cube is going to be stored in the Database ( meaning the in the table space where OLAP is Stored).
3. What is the difference between Analtyic work space manager and discoverer For Olap.
Help is higly appreciated ..
Thanks, Prasad

One thing to point out : The OLAP option is a costed database option, so while this is automatically installed and part of your database you do have to purchase additional licenses to use this feature. You will need to check with your Oracle account manager to see if you are actually licensed to use this feature.
If you have existing 9i OLAP cubes these can be quickly and easily migrated to 10g OLAP and the documentation explains how to do this. However, there are many new features that are part of 10g OLAP that will improve the performance of your data model that will not be enabled as part of a migration process. Based on my experiences it would be quicker and easier (depending on the size of your existing 9i OLAP cubes) to consider rebuilding your data model using these new 10g features. You should be able to export all the dimensions to XML templates from 9i OLAP and import the templates into the 10g schema and reload your dimensions.
For the cubes you will probably want to consider using partitioning, composites and compression to provide maximum flexibility and performance for your new data model. There is more information on these features within the OLAP documentation and in the many whitepapers and presentations on the OLAP home page on OTN.
For moving data from SQL Server to 10g OLAP much depends on the nature of the data transfer. If it is a one-off bulk data load then you could consider using the normal MS command line tools to dump the data out and transfer it to Oracle. Alternatively, you could consider using Oracle Data Integrator to manage the extraction of the SQL Server data and the data load process into Oracle 10g relational tables. If you decide to use Data Integrator this will require additional licenses.
If you can extract the data from SQL Server to flat files you can use Oracle Warehouse Builder (basic ETL is free as part of the 10g database license) to load that flat file data via external tables. Warehouse Builder can also be used to define your OLAP data model (think of Warehouse Builder as a more powerful version of Analytic Workspace Manager) and provides tools to load the data directly into your OLAP dimensions and cubes. You can get more information on managing OLAP data models with Warehouse Builder the Warehouse Builder home page on OTN.
There are no facilities to modify Discoverer Viewer to add customisations. You can add company logos, modify certain colors and/or hide certain features but it is not possible to add additional features. You may want to consider using BI Beans to provide this type of environment. BI Beans is the development framework used to create Discoverer Viewer. You can create customised JSP pages that look identical to Discoverer Viewer pages with the added benefit of providing your own specific features. You can get more information on the BI Beans home page on OTN.
Keith Laker
Oracle EMEA Consulting
BI Blog: http://oraclebi.blogspot.com/
DM Blog: http://oracledmt.blogspot.com/
BI on Oracle: http://www.oracle.com/bi/
BI on OTN: http://www.oracle.com/technology/products/bi/
BI Samples: http://www.oracle.com/technology/products/bi/samples/

Oracle Per5formance Issue With Huge Data

Hi Everyone,
I needed one help from the techies here.
The issue I am facing is, I have got a set of oracle queries that I needed to execute often.
Previously the things were fine as the number of records were less.
But now I have gone into deep trouble as the new dump I have got is having tremendous amount of data.
For example one of the table my query used has got 8000000 records.
Can anyone tell me about the ways that I can use to improve the perforance level.
Most of the tables i have already seen are using indexes.
Thanks in Advance

user12280955 wrote:
For example one of the table my query used has got 8000000 records.So? The number of rows does not determine performance.
SQL> set timing on
SQL> select count(*) from daily_xxxxx;
COUNT(*)
2332931003
Elapsed: 00:00:09.94Approx 1,125,793 I/O calls (majority logical and not physical) were needed and performed in 10 seconds.
Compare this to the following:
SQL> select count(*) from all_objects;
COUNT(*)
72873
Elapsed: 00:00:10.29A tiny percentage of rows in comparison with the 1st query above. 1,473,899 I/O calls make (majority also logical).
What are the common factors here? Elapsed execution time and the amount of I/O done. The execution time has nothing to do with the number of rows processed.
So if number of rows are effecting your database's performance, then your database design and physical implementation is not optimal - as it does not provide optimal I/O paths to the desired result sets that need to be determined by the database. In other words, the design is not scalable. Relational design is inherently scalable - when done correctly.
So I would think you need to review your design - both logical and physical.

RDDGENOL job getting cancelled with huge number of ABAP dumps

Hi,
In our Production system, we had applied Kernel upto patch level 263 and while
applying SPAM 30 for release 640 it failed. We tried applying Basis patch 15 still
it did not went through.
Since then we are getting around 1500-2500 dumps daily. The job RDDGENOLis getting cancelled and the dumps RAISE_EXCEPTION and SYNTAX_ERROR is displayed in ST22. Today, there were around 2500 dumps in the system.
When I now run the Database <-> ABAP Dictionary consistency check in
DB02, I see number of errors for missing objects in the database. There
are around of dozens of Primary Indexes, secondary Indexes, Tables,
Views listed in the result:
Objects Missing in database No.
=========================== ===
Primary indexes 6
Secondary indexes 14
Tables 37
Views 26
When I try to create each of them using SE16 or using SQL utility on OS
level, it gives an SQL system error. We are also getting thousands of
dumps each day (attached alongwith this message). Also I see the CPF327E error
in the OS Level Joblog.
Also we cannot transport any requests. There is some problem with or
due to the conversions. Is it a problem with Kernel, ABAP dictionary or
database.
Following is the extract of the dumps due to DDIC in 000:
============================================================
Runtime Error          SYNTAX_ERROR
Date and Time
ShrtText
Syntax error in program "CL_WB_CROSSREFERENCE==========CP ".
What happened?
Error in ABAP application program.
The current ABAP program "SAPLSEWB" had to be terminated because one of the
statements could not be executed.
This is probably due to an error in the ABAP program.
In program "CL_WB_CROSSREFERENCE==========CP ", the following syntax error
occurred
in the Include "CL_WB_CROSSREFERENCE==========CM00D " in line 8:
|    "The type "KEYWORD_TAB" is unknown."
============================================================
============================================================
Runtime Error          RAISE_EXCEPTION
Date and Time
ShrtText
Exception condition "DESTINATION_NOT_OPEN" raised.
What happened?
The current ABAP/4 program encountered an unexpected
situation.
What can you do?
Print out the error message (using the "Print" function)
and make a note of the actions and input that caused the
error.
To resolve the problem, contact your SAP system administrator.
You can use transaction ST22 (ABAP Dump Analysis) to view and administer
termination messages, especially those beyond their normal deletion
date.
is especially useful if you want to keep a particular message.
Error analysis
A RAISE statement in the program "SAPLCRFC" raised the exception
condition "DESTINATION_NOT_OPEN".
Since the exception was not intercepted by a superior program
in the hierarchy, processing was terminated.
Short description of exception condition:
Specified destination is not open.
For detailed documentation of the exception condition, use
Transaction SE37 (Function Library). You can take the called
function module from the display of active calls.
How to correct the error
You may able to find an interim solution to the problem
in the SAP note system. If you have access to the note system yourself,
use the following search criteria:
"RAISE_EXCEPTION" C
"SAPLCRFC" or "LCRFCU20"
"RFC_CONNECTION_CLOSE"
or
"SAPLCRFC" "DESTINATION_NOT_OPEN"
or
"RADBAT01 " "DESTINATION_NOT_OPEN"
If you cannot solve the problem yourself and you wish to send
an error message to SAP, include the following documents:
1. A printout of the problem description (short dump)
To obtain this, select in the current display "System->List->
Save->Local File (unconverted)".
2. A suitable printout of the system log
To obtain this, call the system log through transaction SM21.
Limit the time interval to 10 minutes before and 5 minutes
after the short dump. In the display, then select the function
"System->List->Save->Local File (unconverted)".
3. If the programs are your own programs or modified SAP programs,
supply the source code.
To do this, select the Editor function "Further Utilities->
Upload/Download->Download".
4. Details regarding the conditions under which the error occurred
or which actions and input led to the error.
System environment
SAP Release.............. "640"
Application server....... "TXTPDSAP"
Network address.......... "172.16.0.140"
Operating system......... "OS400"
Release.................. "5.3"
Hardware type............ "0065000655EC"
Character length......... 8 Bits
Pointer length........... 64 Bits
Work process number...... 35
Short dump setting....... "full"
Database server.......... "TXTPDSAP"
Database type............ "DB400"
Database name............ "TXT"
Database owner........... "R3TXTDATA"
Character set............ "en_US.ISO8859-1"
SAP kernel............... "640"
Created on............... "Dec 11 2008 23:06:45"
Created in............... "AIX 1 5 00538A4A4C00 (IBM iSeries with OS400)"
Database version......... "DB4_52"
Patch level.............. "263"
Patch text............... " "
Supported environment....
Database................. "V5R2, V5R3, V5R4, V6R1"
SAP database version..... "640"
Operating system......... "OS400 2 5, OS400 3 5, OS400 4 5, OS400 1 6"
Memory usage.............
Roll..................... 696832
EM....................... 16759712
Heap..................... 0
Page..................... 32768
MM Used.................. 1383120
MM Free.................. 3483600
SAP Release.............. "640"
User and Transaction
Client.............. 000
User................ "DDIC"
Language key........ "E"
Transaction......... " "
Program............. "SAPLCRFC"
Screen.............. "SAPMSSY0 1000"
Screen line......... 6
Information on where terminated
The termination occurred in the ABAP program "SAPLCRFC" in
"RFC_CONNECTION_CLOSE".
The main program was "RADBAT01 ".
The termination occurred in line 22 of the source code of the (Include)
program "LCRFCU20"
of the source code of program "LCRFCU20" (when calling the editor 220).
The program "SAPLCRFC" was started as a background job.
Job name........ "RDDGEN0L"
Job initiator... "DDIC"
|    Job number...... 00032101
============================================================
Kindly let me know the immediate fix. Thanks!
Regards,
Nick

Hi Nick,
I would say, you are having a slight missperception of this forum ... it should not replace your basis ressources in your company.
What you are doing is pretty complex (and the mixtrure of SPAM Update AND Basis SP seems to be wrong to me) and therefore, it will not be possible, to help you, just based on such a short dump.
Sorry,
Volker Gueldenpfennig, consolut international ag
http://www.consolut.net - http://www.4soi.de - http://www.easymarketplace.de

BPC Performance with large number of dimensions's members

Hi,
I would like to know if there is a limitation on the number of members in one dimension. This dimension, named PROJET, is often used in expansion on our input schedule reports (to retrieve the projects which belong to the entity entered in the current view).
With approximate 2500 members in this dimension, the report takes about 4 minutes to be expansed (or even open).
We have 8 dimensions with few to 300 members. The PROJECT dimension is the biggest in terms of members's number.
Thank you in advance for your feedback !
Helene

Hi Helene,
With 3,000 members you should not be experiencing these problems, if your report is designed properly.
I'm running BPC 5.1 SP8 on SQL 2005, with a dimension containing 22,000 members. Client PC's are typical (XP, Excel 2007, 1 or 2 gig RAM).
Using EVDRE and this dimension expanding on the rows, most reports & input schedules can expand & refresh in the range of 10 to 30 seconds.
The faster times are when I use a row expansion memberset using dimension properties, such as Active="X". The slower times are when the memberset is hierarchy-based, such as BAS.
If you're using a dynamic template (one using EVEXP for the expansion) then you should start over using EVDRE. It will be much faster, particularly if you optimize your row expansion.
It's often a good idea to add dimension properties specifically for the purpose of optimizing the report expansion, if the dimension has thousands of members. I sometimes go as far as to add properties which mimic the hierarchy (MyLevel2, MyLevel3, etc) just for this purpose.

Error in deploying an OLAP cube with OWB ---- Oracle ACE's , please help me

I am using OWB 10.2 for building a BI application . All goes well until I tried to deploy an OLAP relational cube . When I cofigure an OLAP cube -deplyoment options- with 'deploy all' value ie OWB genarete un pl/sql script with cmw2 procedures , Control Center Manager can't deploy my cube and give me this log:
ORA-06510 : PL/SQL unhandled user-defined exception
ORA-06512 : at OLAPSYS.CMW2_OLAP_UTILITY , line 1660
ORA-01403 : no data found
ORA-06512 : at "OLAPSYS.CMW2_OLAP_CUBE, line 33
ORA-06512 : at "OLAPSYS.CMW2_OLAP_CUBE, line 55
ORA-06512 : at "OLAPSYS.CMW2_OLAP_CUBE, line 386
ORA-06512 : at line 5
What do you think is wrong?
TIA

Ragnar is quite correct. When you design a cube there is an associated table generated and FK constraints are created that link back to the related dimension objects and their associated tables. The OLAP dimensions and cubes are just pieces of metadata and don't actually exist as first class objects in their own right. Most importantly don't confuse OLAP dimensions with relational dimensions that are created to support query re-write.
In Control Center if you check the Table node in your tree you should see a list of tables that are the relational source objects for your dimensions and cubes. Deploy the dimension tables first, then deploy the dimensions using the "Deploy All" option. Lastly depoly your table followed by the cube.
In Control Center it is possible to force OWB to deploy all related objects. I think it is on one of the menu options within Design Center. I will see if I can find the exact details when I next have OWB open.
Hope this helps
Keith Laker
Oracle EMEA Consulting
BI Blog: http://oraclebi.blogspot.com/
DM Blog: http://oracledmt.blogspot.com/
BI on Oracle: http://www.oracle.com/bi/
BI on OTN: http://www.oracle.com/technology/products/bi/
BI Samples: http://www.oracle.com/technology/products/bi/samples/

Oracle OLAP as OBIEE Data Source

I've got a couple of questions regarding the use of Oracle OLAP (Analytic Workspace/Cube) as an OBIEE data source.
First: As a general rule when creating a dimension, we create a total roll-up for the dimension i.e. "Total Product", "Total Geog", "Total Customer" etc... Generally, I don't create a total roll-up for time dimensions. When importing metadata from OLAP to OBIEE, OBIEE creates a "Total" level for all dimensions. Now, I understand why OBIEE does that; to support queries that might exclude one or more dimensions. My question is: what is the best method/procedure to deal with the extra "Total" level?
Second: I would appreciate it if someone could explain this error for me: [nQSError: 59137] Filter level YEAR is below the projected level Total on dimension CMP_TIME while an externally aggregated measure is present. (HY000). I understand the words, but have no clue what OBIEE is trying to tell me. This error pops up constantly and I see no rhyme or reason that would cause it. the specific case above occurred when I clicked on the sort icon for a measure included in a report.
Thanks,

Mark,
Thanks for the reply. However, I'm not sure I made myself clear. I have created a "Product" dimension in AWM (Analytic Workspace Manager) in the following structure: Product -> Product Line -> Total Product. Withing the context this hierarchy, "Total Product" is the "Grand Total" Level. When this data is imported into OBIEE using "Oracle OLAP" as a data source, the Product hierarchy is created in the Physical Layer as an "Oracle OLAP Dimension". In the BMM Layer, the hierarchy is structured as: Product -> Product Line -> Total Product -> Total. There are now two "Total" Levels. Naturally only one, the OBIEE generated Total, is defined as "Grand Total". The only child of the Total level is Total Product. I have two hierarchy levels that are the same. So, do we need both? should we keep both? Should a dimension defined within AWM for use in OBIEE NOT include a total level? It's not really a problem, it just doesn't seem to make any sense to have TWO total levels within a hierarchy.
On the second issue, I wish I could provide some detail, but I'm really not sure how I'd do that. That's why I asked for the meaning of the error. What is OBIEE telling me that I'm doing wrong. All I really did was import the metadata, drag it to the BMM Layer, deleted some of the hierarchy level keys, renamed some columns and dragged the stuff over to the Presentation Layer. So, it's pretty much drag-and-drop.
Another example of the error: We have a Category Dimension (Sub Category -> Category -> Category Group -> Model -> All Categories -> Total) and I want to see the top 10 values of a measure by Category by Model. In an Analysis, adding the Model column works fine, just not the best visualization. Move the Model column to "Sections" and all works; move the Model column to Pivot Table Prompts and it errors. Obviously, I'm asking OBIEE to do something it doesn't want to do, so I'm looking for the root cause of the error.
Thanks,

Maximum number of members in a dimension

Hi,
Is there a limitation of number of members in a dimension in Oracle Olap?
Thanks,
Krishna

I think you got answer for this, which I'll post here for the benefit of other viewers:
"There is no hard limit on the number of members of dimension in Oracle OLAP. You may run into some practical limits based on your HW and cube design, but that won’t be until you’re dealing with dimensions in the multimillion member range."
Say you have a dimensions of 2 million values and another 3 dimensions with 1000 values each. Potential size of the cube if full of data is 2,000,000 * 1000*1000*1000 which make it big. Say you have a couple more dimensions 10,000 and 10,000 ..even bigger.
Compressing the cube to make sure its as small as possible, where patterns of null cells are not stored is good practice most of the time, especially great when there are lots of null values when there's large dimensions. Partitioning and general good modelling are also key. See the OLAP User's Guide for more information.

Oracle DB 10g and BI Discoverer for OLAP - Dimension Attributes

Hi,
We are using Oracle Database 10g release 1 with partitioning, data mining and OLAP options and Analytic Workspace Manager 10.2.0.1.0A to create the multidimensional objects. For the user dimensions created using AWM we have custom attributes like HireDate, StartDate, Sales Personnel Role etc., For reporting purposes we are using Discoverer for OLAP. In this Discoverer version, I don't see an explicit provision to drag these attributes onto the worksheet. We are only able to filter based on these attributes and capture the measures...
Can someone throw light on this? Also, if there is a possibility to drag these attributes onto the worksheet can that be expounded?
Thanks in advance!

Again this depends on what you are trying to achieve. If you define an attribute against a dimension it takes very little space as it is not directly connected to a cube and so no data is stored against that attribute.
If, however, you have a 4D revenue cube (product, geography, channel, time) with products attributes COLOR and PACK SIZE and you want to view revenue additionally borken-out by COLOR and PACK SIZE as well as the other four dimensions then your schema will require additional storage space. However, 10g compressed cubes and sparsity options do help to manage the explosion of data points as the number of dimensions increases. This should allow you to easily add attributes as dimensions into your cube.
One thing to remember is that most users start to struggle when confronted with more than 9 dimensions. So although Oracle OLAP can create extremely large dimensional models, users prefer their cubes to have 9 or fewer dimensions.
Hope this helps,
Keith
Oracle Business Intelligence Product Management
BI on Oracle: http://www.oracle.com/bi/
BI on OTN: http://www.oracle.com/technology/products/bi/
BI Beans http://www.oracle.com/technology/products/bib/index.html
Discoverer: http://www.oracle.com/technology/products/discoverer/
BI Software: http://www.oracle.com/technology/software/products/ias/devuse.html
Documentation: http://www.oracle.com/technology/documentation/appserver1012.html
BI Samples: http://www.oracle.com/technology/products/bi/samples/
Blog: http://oraclebi.blogspot.com/

Error in Running a dimension ( BIB-9509 Oracle OLAP did not create cursor.)

oracle.dss.dataSource.common.QueryRuntimeException: BIB-9509 Oracle OLAP did not create cursor.
oracle.express.ExpressServerExceptionError class: OLAPI
Server error descriptions:
DPR: Unable to create server cursor, Generic at TxsOqDefinitionManagerSince9202::crtCurMgrs4
OES: ORA-00938: not enough arguments for function
, Generic at TxsRdbRandomAccessQuery::TxsRdbRandomAccessQuery
help is appreciated
Thanks, Prasad

This is the patch: 2529822
"Best Practices for Tabular Cube Aggregation & Query Operations". Thanks very much for your advice.
I followed the instructions in the document, but it did not seem to make a difference when
trying to follow the tutorial instructions and using the SALES measure.
However, I selected the stock price measures and it worked ok.
Is there a size limit here? I am looking at Oracle OLAP and BIBeans to replace our Cognos installation,
which is expensive and unreliable on unix. Of course our managers and customers want every dimension
across a huge fact table to be available for end users...

Huge number of idle connections from loopback ip on oracle RAC node

Hi,
We have a 2node 11gR2(11.2.0.3) oracle RAC node. We are seeing huge number of idle connection(more than 5000 in each node) on both the nodes and increasing day by day. All the idle connections are from VIP and loopback address(127.0.0.1.47971 )
netstat -an |grep -i idle|more
127.0.0.1.47971 Idle
any insight will be helpful.
The server is suffering memory issues occasionally (once in a month).
ORA-27300: OS system dependent operation:fork failed with status: 11
ORA-27301: OS failure message: Resource temporarily unavailable
Thanks

user12959884 wrote:
Hi,
We have a 2node 11gR2(11.2.0.3) oracle RAC node. We are seeing huge number of idle connection(more than 5000 in each node) on both the nodes and increasing day by day. All the idle connections are from VIP and loopback address(127.0.0.1.47971 )
netstat -an |grep -i idle|more
127.0.0.1.47971 Idle
any insight will be helpful.
The server is suffering memory issues occasionally (once in a month).
ORA-27300: OS system dependent operation:fork failed with status: 11
ORA-27301: OS failure message: Resource temporarily unavailable
Thankswe can not control what occurs on your DB Server.
How do I ask a question on the forums?
SQL and PL/SQL FAQ
post results from following SQL
SELECT * FROM V$VERSION;

Maximum number of members in a dimension (BSO)

we are designing a new planning application and the requirement for one of the dimensions may force us to have a large number of members. I needed a ballpark figrue on what the maximum number of members should be in planning for it to work in an acceptable manner (aggregate times, run time calculations etc) ... the number we are looking at is over 30,000 members and another 15,000 shared members 10-11 levels deep. If any one has experience with such large dimension, can you shed some light on how long the aggregation times would take and if we use it in form, how best to agg the RTPs for that dimension? thank you in advance.

Check out below posts:
For facts:
Members in Essbase
Maximum members in a Dimension
For limits:
Maximum members for a Dimension in a BSO cube
Cheers..!!

Oracle OLAP:Same member values across dimensions

We had an enquiry from a client which mentions:
Per Essbase the same dimension member value cannot exist across multiple dimensions. e.g. if we have a account dimension with member 1001 , it is not possible to have product dimension with same member value 1001. This is essentially because Essbase implements it like a shared member. Apparently Essbase Excel-addin will not know what dimension to search for if both have same values.
As per our understanding there is no such limitation in Oracle OLAP both during cube build or querying via Excel Add-in. i.e we can have account dimension with member 1001 and same member value can exist in a product dimension. Please advise and if anybody has any reference to any documentation in this regard please let us know.

Hi there,
You are correct - there is no such limitation in Oracle OLAP.
A dimension value must be unique in any given dimension but can be repeated in other dimensions.
I do not have any documentation but this is really easy to prove using the OLAP worksheet in AWM, eg:
aw create test
define dim1 dimension text
maintain dim1 add 'MEMBER1'
report dim1
define dim2 dimension text
maintain dim2 add 'MEMBER1'
report dim2
aw detach test
aw delete testI hope this helps
Thanks,
Stuart Bunby
OLAP Blog: http://oracleOLAP.blogspot.com
OLAP Wiki: http://wiki.oracle.com/page/Oracle+OLAP+Option
OLAP on OTN: http://www.oracle.com/technology/products/bi/olap/index.html
DW on OTN : http://www.oracle.com/technology/products/bi/db/11g/index.html

Physical size of dimensions and composites (Express and Oracle OLAP)

Hi, have a rather odd question re: the physical size (in bytes) of composites...are they affected by the dimension sizes?
I.e. if I define a dimension "Org" as:
define org dimension text width 10
will that result in smaller composites than if I define it as
define org dimension text width 30
? i.e. given that everything else if equal (same number of values, same # of tuples, etc.) does the physical definition of how wide a dimension is affect the size of the composites that use that dimension?
p.s. need to know this for Express, but if Oracle OLAP doesn't behave the same I'd like to know that also.
Thanks!
Scott

A 5 dimensional composite has keys built of 5 4-byte dimension positions. There is a copy of the 20 byte value stored indexed by physical position that we use internally to determine the value of the base dimensions. Also there is a copy stored in a hash table used to take the base dimension values and convert those back to a physical position (that's 20 bytes for the key and 4 bytes for the physical position).
Finally, we have internal structures ("inversions") that we use to essentially do "limit composite to base dimensions" for looping (for each base dimension value, a list of physical positions that have that base value in it), at 4 bytes per dimension or another 20 bytes.
That's 64 bytes. Yeah, it's not cheap, but compared to a dense variable, you save a ton of space. I seem to remember a calculation someone did a long time ago that said that composites were net smaller for anything less than 20% dense.
I'm not surprised that under certain conditions you can find a better data layout. Certainly the overhead of storing dense data is tiny compared to a composite. If you are clever (or lucky) in your segmenting, you can take advantage of the "holes" in the data to improve the storage. The "holes" (large groups of NA values) can go for many pages, if the faster varying dimension size is large, and won't be stored.
Oracle OLAP contains various improvements over the Express code. I believe we now use compressed bitmaps to store the inversions, rather than lists of values, for a substantial savings. There may be other changes as well (I'm not current on the details of the dimension storage).
Hope this helps.
Jim

Dimension table is associated with N number of fact tables

Hi Expert's,
In case of star schema one fact table is associated with N dimension tables , is it possible to join one dimension table with N number of fact tables...please let my know the your suggestion...
Thank's
Harish

is it possible to join one dimension table with N number of fact tablesYes you can join one dimension table to many fact tables.
Thanks,
Vino

ORACLE olap - dimension with huge number of members.

Similar Messages

Maybe you are looking for