Oracle® Business Intelligence Applications Installation Guide for Informatica PowerCenter Users Release 7.9.6.4 Part Number E35271-01 |
|
|
PDF · Mobi · ePub |
This section provides information about preparing to install and deploy Oracle BI Applications. Oracle recommends that you first refer to Section 3.1, "General Guidelines for Setting Up Databases for Oracle BI Applications", and then refer to the sections that apply to the database type that you are using.
Note:
For the most up-to-date information about supported databases and source systems in this version of Oracle Business Intelligence Applications, make sure you read System Requirements and Supported Platforms for Oracle Business Intelligence Applications. Make sure that you also read the Oracle Business Intelligence Applications Release Notes. The most up-to-date versions of these documents are located on the Oracle Technology Network at http://www.oracle.com/technetwork/middleware/bi-foundation/documentation/bi-apps-098545.html
. To register for a free account on the Oracle Technology Network, go to http://www.oracle.com/technetwork/index.html
.
You should review this information before you begin the installation and deployment process. You should also read the general guidelines for setting up the Oracle Business Analytics Warehouse and read the appropriate database-specific guidelines for the source OLTP databases that you are using.
You also need to satisfy the database and Informatica PowerCenter requirements that are specified in Section 4.3, "Mandatory Requirements."
For information about database-specific settings, see System Requirements and Supported Platforms for Oracle Business Intelligence Applications.
For information about code page settings, refer to the Informatica documentation.
This section contains the following topics:
Section 3.1, "General Guidelines for Setting Up Databases for Oracle BI Applications"
Section 3.2, "Partitioning Guidelines For Large Fact Tables"
Section 3.3, "Miscellaneous Information About Oracle BI Applications Deployments"
Section 3.4, "Oracle-Specific Database Guidelines for Oracle Business Analytics Warehouse"
Section 3.5, "IBM DB2 UDB-Specific Database Guidelines for Oracle Business Analytics Warehouse"
Section 3.7, "SQL Server-Specific Database Guidelines for Oracle Business Analytics Warehouse"
Section 3.8, "Teradata-Specific Database Guidelines for Oracle Business Analytics Warehouse"
This section contains general guidelines for source and target databases, and contains the following topics:
Section 3.1.2, "General guidelines for Oracle Business Analytics Warehouse"
Section 3.1.3, "Why Use a Separate Database for Oracle Business Analytics Warehouse?"
Section 3.1.4, "Miscellaneous Information About Oracle BI Applications Databases"
General guidelines for transactional databases:
On the transactional database, you should place the S_ETL tables in a separate tablespace. These ETL tables are used by the Oracle Business Analytics Warehouse and should not be part of the routine backup processes.
A complete listing of these tables is available in Oracle Business Analytics Warehouse Data Model Reference.
To maximize ETL performance for Siebel CRM source systems running on DB2 databases, create three indexes on the Siebel OLTP database, using the following SQL commands:
CREATE INDEX S_AUDIT_ITEM_M100 ON S_AUDIT_ITEM (FIELD_NAME ASC, BUSCOMP_NAME DESC) PCTFREE 10 ALLOW REVERSE SCANS COLLECT DETAILED STATISTICS;
CREATE INDEX S_AUDIT_ITEM_M101 ON S_AUDIT_ITEM (RECORD_ID ASC, FIELD_NAME DESC) PCTFREE 10 ALLOW REVERSE SCANS COLLECT DETAILED STATISTICS;
CREATE INDEX S_OPTY_M102 ON S_OPTY (ROW_ID ASC, PR_POSTN_ID DESC) PCTFREE 10 ALLOW REVERSE SCANS COLLECT DETAILED STATISTICS;
The following general guidelines will help you set up the data warehouse physical database for performance and growth:
At a minimum, separate the data and index tablespaces. Create more tablespaces to separate heavily used tables and their indexes.
Use the maximum block and page size available for tablespaces (for example, 32K), because it provides good overall performance and also does not impose low limits to the maximum size to which the tablespace can grow, as compared to 4K, 8K,16K sizes.
If you are using multiple disk storage systems, stripe the tablespace containers and files across as many disks as possible.
Raw devices for tablespaces provide better performance as compared to cooked file systems.
RAID-5 is known to give a good balance of performance and availability.
For Oracle databases, size the buffer pools based on content and size (number of tables and their sizes) of tablespaces.
Allocate about 75 percent of the total available server memory to the database, assuming no other application is running on the same server.
During the Oracle Business Analytics Warehouse configuration process, when you create the data warehouse tables using the procedure described in Section 4.10, "Setup Task: Creating the Oracle Business Analytics Warehouse Tables," you can create tables in one tablespace and indexes in another tablespace. However, for performance reasons, it is recommended that you create tablespaces as described in Table 3-1.
Table 3-1 Recommended Tablespace Configuration
Tablespace Name | List of Tables |
---|---|
DIM_STG |
W_*DS |
FACT_STG |
W_*FS |
DIM |
W_*D and W_*MD |
FACT |
W_*F |
AGG |
W_*A |
OTHER |
Remaining W* tables |
DIM_INDX |
Indexes of W_*D tables (for example, other tables would include W*G and W*GS tables) |
FACT_INDX |
Indexes of W_*F tables |
OTHER_INDX |
Remaining indexes of W* tables |
Note:
To avoid fatal deadlocks during the ETL, make sure that you select the 'Session Level Retry on Deadlock' option in Informatica.
Although it is technically possible to put the Oracle Business Analytics Warehouse in the same database as the transactional database, it is not recommended for performance reasons. The transactional database is structured as an online transaction processing (OLTP) database, whereas the Oracle Business Analytics Warehouse is structured as an online analytical processing (OLAP) database, each optimized for its own purpose. The reasons for not combining the two databases are:
The analytical queries interfere with normal use of the transactional database, which is entering and managing individual transactions.
The data in a transactional database is normalized for update efficiency. Transactional queries join several normalized tables and will be slow (as opposed to pre-joined, de-normalized analytical tables).
Historical data cannot be purged from a transactional database, even if not required for current transaction processing, because you need it for analysis. (By contrast, the analytical database is the warehouse for historical as well as current data.) This causes the transactional database to further slow down.
Transactional databases are tuned for one specific application, and it is not productive to use these separate transactional databases for analytical queries that usually span more than one functional application.
The analytical database can be specifically tuned for the analytical queries and Extract-Transform-Load (ETL) processing. These are quite different from transactional database requirements.
The Informatica Repository stores all of the Informatica object definitions for the ETL mappings that populate the Oracle Business Analytics Warehouse. It is a series of repository tables that are stored in a database, which can be a transactional, analytical, or separate database.
Oracle Business Analytics Warehouse works with relational database management systems. In addition to the general requirements, there are additional database management systems (DBMS)-specific requirements depending on the DBMS you are using.
For information about partitioning fact tables for Oracle Business Intelligence Applications, see "Oracle Business Intelligence Applications Version 7.9.6.x Performance Recommendations [ID 870314.1]" on My Oracle Support.
This section contains the following topics:
Section 3.3.1, "Preconfigured Mapping for the SA System Subject Area"
Section 3.3.3, "Creating Custom Indexes in Siebel Source Databases for Incremental Load Performance"
Section 3.3.5, "Running an ETL with a Non-English OLTP Data Source"
For Oracle's Siebel Applications customers, Table 3-2 describes the preconfigured mappings for the SA system subject area. Fields that are not available in Oracle's Siebel transactional database will default to values shown in the table.
Overriding Defaults. You can add user-specific values for these fields, by creating an extension table to the S_USER table. to store the user-specific defaults for these fields. Additionally, you can change any of the default values. The metadata for the following logical table can be modified to include any physical extension table.
SA User.(User)
For instructions, refer to the documentation about configuring tables and columns for Oracle's Siebel Business Applications.
Setting Provider Information. Typically, the cell phone and the fax numbers in the Oracle Business Analytics Warehouse do not contain a provider name. Therefore, the Pager will typically be a numeric value such as 555-483-3843. To append a provider to this address, use the following guidelines:
If the entire company has the same provider, then you can append the provider in the column mappings.
If users can have different providers, you need to create an extension table. For instructions, refer to the documentation about configuring tables and columns for Oracle's Siebel business applications.
Table 3-2 Preconfigured Mappings for the User Table in the SA System Subject Area
Logical Column | Physical Table | Expression | Comments |
---|---|---|---|
Cell Phone |
'' |
It might be mapped to S_CONTACT.CELL_PH_NUM if this field contains SMTP address. |
|
Cell Phone Priority |
'' |
Defaults to N |
|
Display Name |
S_CONTACT |
"Real Time OLTP"."".SIEBEL.S_CONTACT_User.FST_NAME || ' ' || "Real Time OLTP"."".SIEBEL.S_CONTACT_User.LAST_NAME |
First Name concatenated with Last Name |
|
S_CONTACT |
EMAIL_ADDR |
|
Email Priority |
'HNL' |
Defaults to N |
|
Email Type |
'html' |
Defaults to HTML |
|
Group Name |
S_RESP |
NAME |
|
Handheld |
'' |
Defaults to an empty string |
|
Handheld Priority |
'' |
Defaults to an empty string |
|
Language |
'en' |
Defaults to 'en' |
|
Locale |
'en' |
Defaults to 'en' |
|
Logon |
S_USER |
LOGIN |
|
Pager |
'' |
It could be mapped to S_CONTACT.PAGER_PH_NUM if this field contains SMTP address |
|
Pager Priority |
'' |
Defaults to N |
|
Time Zone |
S_TIMEZONE |
NAME |
Table 3-3 lists some of the initialization blocks common to all Oracle BI Applications and their purposes. Initialization blocks that are specific to each Oracle BI Applications area are not listed here.
To view the initialization blocks provided with Oracle BI Applications, open the Variable Manager in the Oracle Business Intelligence Enterprise Edition Administration Tool. For instructions, see Oracle Fusion Middleware System Administrator's Guide for Oracle Business Intelligence Enterprise Edition.
Table 3-3 Initialization Blocks and Their Purposes
Initialization Block | Purpose |
---|---|
Authorization |
Calculates user's responsibilities from the database. |
Authentication |
Authenticates against the database and verifies the user exists as a operational application user. |
External Metadata Strings |
Gets the translated value of metadata strings for the user's locale. This initialization block is critical to Intelligence Dashboards in international deployment. |
LOCALE |
Sets the user's locale specification in the Oracle BI Server. |
Login Properties |
Gets user's login properties, such as full name, from the database. This initialization block also sets the logging level of the users. By default, the log level is 0 for all users. If you wish to generate Oracle BI query logs for all users, this initialization block should be updated by changing the default value as well as the value in the initialization SQL. |
Default System Language ID |
Sets the variable OLTP_LANG_ID by querying the operational application database. |
Organizations for Org-based Security |
Queries the operational application database to get the organization membership for each user. It sets the variable ORGANIZATION. |
Primary Owner ID |
Sets the primary owner ID based on the user login ID. |
Primary Position ID |
Queries the operational application database to set the variable PRIMARY_POSTN_ID. |
Warehouse Refresh Date |
Sets several time-based variables such as CURRENT_YEAR. |
ETL Run Date |
Retrieves the ETL run date. |
ETL Default Currency |
Retrieves the default currency. |
To maximize performance for Siebel CRM, you can implement indexes using SQL files that are available in the ORACLE_HOME\biapps\dwrep directory. Table 3-4 describes the SQL files that are appropriate for the specified applications.
Table 3-4 SQL files for Siebel transactional databases
Application Name | SQL File Name |
---|---|
Horizontal Application |
|
Industry Application |
The SQL files generate indexes on all S_.* tables that are being used by the preconfigured applications.
Note:
If you move from a test to a production environment, you need to drop the indexes and re-create them in the production environment.
Change capture SQL generates the following SQL:
Insert into S_ETL_I_IMG_XX (ROW_ID, LAST_UPD) AS SELECT ROW_ID, LAST_UPD, MODIFICATION_NUM From S_XXX WHERE LAST_UPD > 'LAST REFRESH_DATE – PRUNE DAYS' AND NOT EXISTS ( SELECT 'X' FROM S_ETL_R_IMAGE WHERE S_ETL_R_IMAGE.ROW_ID = S_XXX.ROW_ID AND S_ETL_R_IMG_XX.MODIFICATION_NUM = S_XXX.MODIFICATION_NUM AND S_ETL_R_IMG_XX.LAST_UPD = S_XXX.LAST_UPD )
Table 3-5 shows the indexes that, based on the SQL above, are created on the S_CONTACT table by the SQL generation scripts.
Oracle EBS source database tables contain mandatory LAST_UPDATE_DATE columns, which are used by Oracle BI Applications for capturing incremental data changes. Some Oracle EBS source tables used by Oracle BI Applications do not have an index on the LAST_UPDATE_DATE column because the presence of the index can impede performance of the source application.
There are three categories of Oracle EBS tables as they relate to indexes on the LAST_UPDATE_DATE column:
Category 1: Tables that do not have indexes on the LAST_UPDATE_DATE column but on which indexes can be created without impeding performance.
Category 2: Tables that have indexes on LAST_UPDATE_DATE columns. These indexes were introduced in Oracle EBS release 12.
Category 3: Tables that cannot have indexes on the LAST_UPDATE_DATE column because performance will be impeded in the Oracle EBS environment.
The following DDL script creates custom indexes on the LAST_UPDATE_DATE column for Category 1 tables, that is, tables in all Oracle EBS releases that do not already have this index created and for which there are no known performance implications for creating such indexes.
You should run this DDL script if your source system is Oracle EBS release 11i or release 12 and you have experienced slow incremental extraction mapping performance while implementing specific subject areas.
Note: If your source system is Oracle EBS release 12, Oracle EBS release 11.5.10, Oracle EBS release 11.5.9 or lower and has been migrated to Oracle Applications Tablespace Model (OATM), then replace <IDX_TABLESPACE>
with APPS_TS_TX_IDX
.
The DDL script is as follows:
CREATE index AP.OBIEE_AP_INVOICE_PAYMENTS_ALL ON AP.AP_INVOICE_PAYMENTS_ALL(LAST_UPDATE_DATE) tablespace <IDX_TABLESPACE>;
CREATE index AP.OBIEE_AP_PAYMENT_SCHEDULES_ALL ON AP.AP_PAYMENT_SCHEDULES_ALL(LAST_UPDATE_DATE) tablespace <IDX_TABLESPACE>;
CREATE index AP.OBIEE_AP_INVOICES_ALL ON AP.AP_INVOICES_ALL(LAST_UPDATE_DATE) tablespace <IDX_TABLESPACE>;
CREATE index GL.OBIEE_GL_JE_HEADERS ON GL.GL_JE_HEADERS (LAST_UPDATE_DATE) tablespace <IDX_TABLESPACE>;
CREATE index ONT.OBIEE_OE_ORDER_HEADERS_ALL ON ONT.OE_ORDER_HEADERS_ALL(LAST_UPDATE_DATE) tablespace <IDX_TABLESPACE>;
CREATE index PER.OBIEE_PAY_INPUT_VALUES_F ON PER.PAY_INPUT_VALUES_F (LAST_UPDATE_DATE) tablespace <IDX_TABLESPACE>;
CREATE index PER.OBIEE_PAY_ELEMENT_TYPES_F ON PER.PAY_ELEMENT_TYPES_F (LAST_UPDATE_DATE) tablespace <IDX_TABLESPACE>;
CREATE index PO.OBIEE_RCV_SHIPMENT_LINES ON PO.RCV_SHIPMENT_LINES (LAST_UPDATE_DATE) tablespace <IDX_TABLESPACE>;
CREATE index PO.OBIEE_RCV_SHIPMENT_HEADERS ON PO.RCV_SHIPMENT_HEADERS (LAST_UPDATE_DATE) tablespace <IDX_TABLESPACE>;
CREATE index AR.OBIEE_AR_CASH_RECEIPTS_ALL ON AR.AR_CASH_RECEIPTS_ALL (LAST_UPDATE_DATE) tablespace <IDX_TABLESPACE>;
CREATE index WSH.OBIEE_WSH_DELIVERY_DETAILS ON WSH.WSH_DELIVERY_DETAILS (LAST_UPDATE_DATE) tablespace <IDX_TABLESPACE>;
CREATE index WSH.OBIEE_WSH_NEW_DELIVERIES ON WSH.WSH_NEW_DELIVERIES (LAST_UPDATE_DATE) tablespace <IDX_TABLESPACE>;
Note:
Make sure you use FND_STATS to compute statistics on the newly created indexes and update statistics on newly indexed table columns in the Oracle EBS database.
All indexes created with the DDL in this section have the prefix OBIEE_. This prefix does not follow standard Oracle EBS index naming conventions. Therefore, Autopatch might fail during future upgrades. In such cases, the indexes with the OBIEE_ prefix should be dropped and Autopatch restarted.
The following DDL creates custom indexes on the LAST_UPDATE_DATE column for Category 2 tables, that is, tables for which indexes were introduced for the LAST_UPDATE_DATE column in Oracle release 12.
You should run this DDL script if your source system is Oracle EBS release 11i.
Note: If your source system is Oracle EBS release 11.5.10, Oracle EBS release 11.5.9 or lower and has been migrated to Oracle Applications Tablespace Model (OATM), then replace <IDX_TABLESPACE>
with APPS_TS_TX_IDX
.
The DDL script is as follows:
CREATE index PO.RCV_TRANSACTIONS_N23 ON PO.RCV_TRANSACTIONS (LAST_UPDATE_DATE) INITIAL 4K NEXT 2M MINEXTENTS 1 MAXEXTENTS 50 PCTINCREASE 0 INITRANS 2 MAXTRANS 255 PCTFREE 10 tablespace <IDX_TABLESPACE>;
CREATE index PO.PO_DISTRIBUTIONS_N13 ON PO.PO_DISTRIBUTIONS_ALL (LAST_UPDATE_DATE) INITIAL 4K NEXT 2M MINEXTENTS 1 MAXEXTENTS 50 PCTINCREASE 0 INITRANS 2 MAXTRANS 255 PCTFREE 10 tablespace <IDX_TABLESPACE>;
CREATE index PO.PO_LINE_LOCATIONS_N11 ON PO.PO_LINE_LOCATIONS_ALL (LAST_UPDATE_DATE) INITIAL 4K NEXT 2M MINEXTENTS 1 MAXEXTENTS 50 PCTINCREASE 0 INITRANS 2 MAXTRANS 255 PCTFREE 10 tablespace <IDX_TABLESPACE>;
CREATE index PO.PO_LINES_N10 ON PO.PO_LINES_ALL (LAST_UPDATE_DATE) INITIAL 4K NEXT 4K MINEXTENTS 1 MAXEXTENTS 50 PCTINCREASE 0 INITRANS 2 MAXTRANS 255 PCTFREE 10 tablespace <IDX_TABLESPACE>;
CREATE index PO.PO_REQ_DISTRIBUTIONS_N6 ON PO.PO_REQ_DISTRIBUTIONS_ALL (LAST_UPDATE_DATE) INITIAL 4K NEXT 250K MINEXTENTS 1 MAXEXTENTS 50 PCTINCREASE 0 INITRANS 4 MAXTRANS 255 PCTFREE 10 tablespace <IDX_TABLESPACE>;
CREATE index PO.PO_REQUISITION_LINES_N17 ON PO.PO_REQUISITION_LINES_ALL (LAST_UPDATE_DATE) INITIAL 4K NEXT 250K MINEXTENTS 1 MAXEXTENTS 50 PCTINCREASE 0 INITRANS 4 MAXTRANS 255 PCTFREE 10 tablespace <IDX_TABLESPACE>;
CREATE index PO.PO_HEADERS_N9 ON PO.PO_HEADERS_ALL (LAST_UPDATE_DATE) INITIAL 4K NEXT 1M MINEXTENTS 1 MAXEXTENTS 50 PCTINCREASE 0 INITRANS 2 MAXTRANS 255 PCTFREE 10 tablespace <IDX_TABLESPACE>;
CREATE index PO.PO_REQUISITION_HEADERS_N6 ON PO.PO_REQUISITION_HEADERS_ALL (LAST_UPDATE_DATE) INITIAL 4K NEXT 250K MINEXTENTS 1 MAXEXTENTS 50 PCTINCREASE 0 INITRANS 4 MAXTRANS 255 PCTFREE 10 tablespace <IDX_TABLESPACE>;
CREATE index AR.RA_CUSTOMER_TRX_N14 ON AR.RA_CUSTOMER_TRX_ALL (LAST_UPDATE_DATE) INITIAL 4K NEXT 4M MINEXTENTS 1 MAXEXTENTS 50 PCTINCREASE 0 INITRANS 4 MAXTRANS 255 PCTFREE 10 tablespace <IDX_TABLESPACE>;
Note: Make sure you use FND_STATS to compute statistics on the newly created indexes and update statistics on newly indexed table columns in the Oracle EBS database.
To run an ETL with a non-English OLTP data source, you must make a copy of an appropriate source system container, and configure the language, country, and continent parameters.
To run an ETL with a non-English OLTP data source
In DAC, select File, then New Source System to display the New Source System Container dialog box.
Select the Create as a Copy of Existing Container radio button.
From the Existing Containers drop-down list, select the container that you want to copy, then click OK.
Go to the Design view.
Ensure that you have selected the correct container from the Containers drop-down list.
Select the Source System Parameters tab.
Use the Edit tab below the list of Source System Parameters to change the value of the following parameters in the list:
$$DFLT_LANG (Default Language) – For example, for a Japanese data source, change this value to JPN.
(optional) $$DFLT_COUNTRY (Default Country)
(optional) $$DFLT_CONTINENT (Default Continent)
Tip: To find out the value to specify for the $$DFLT_LANG parameter, issue the query 'select VAL from S_SYS_PREF where SYS_PREF_CD=<ETL value>' against the OLTP database. For example, to find out the default ETL language, issue the following command:
select VAL from S_SYS_PREF where SYS_PREF_CD='ETL Default Language';
Create a new ETL plan for the new source system container, and edit its parameters, as follows:
Click the Execute tab.
Click the Execution Plans subtab.
Click New to create a new blank execution tab and use the subtabs below (for example, Subject Areas, Parameters, Ordered Tasks) to specify the execution plan details.
Click Save.
Click Run Now to run the new ETL plan (or use the Schedule tab to specify when you want to run the new execution plan).
This section contains guidelines that are specific to Oracle databases, and contains the following topics:
To configure the Business Analytics Data Warehouse on Oracle databases more easily, refer to the parameter template files (that is, the init<DB version>.ora files). These files are located in <DRIVE>:\<Oracle BI Applications install directory>\dwrep\Documentation\. A separate parameter template file is provided for each database version. For example, the parameter template file for Oracle 11g databases is named init11g.ora.
The parameter template files provide parameter guidelines based on the cost-based optimizer for Oracle 10g and 11g. Use these guidelines as a starting point. You will need to make changes based on your specific database sizes, data shape, server size (CPU and memory), and type of storage. The database administrator should make changes to the settings based on performance monitoring and tuning.
Copy the appropriate template file into your $ORACLE_HOME/dbs directory. Then, review the recommendations in the template file, and make the changes based on your specific database configuration. The database administrator should make changes to the settings based on performance monitoring and tuning considerations.
Note:
The NLS_LENGTH_SEMANTICS parameter enables you to define byte- or character-length semantics. Oracle BI Applications supports BYTE and CHAR values for this parameter. If you are using MLS characters, then you can add this parameter to the parameter template file for your database version (that is, the init<DB version>.ora file).
This section contains additional suggestions for optimizing performance for Oracle databases.
To avoid ORA-00942 errors with a pre-Oracle 11.1 database, you can either enable the native full outer join implementation for all sessions by setting the parameters below as a system parameter, or disable the FULL_OUTER_JOIN_SUPPORTED feature in the RPD.
alter session set "_optimizer_native_full_outer_join" = 'FORCE'; alter system set "_optimizer_native_full_outer_join" = 'FORCE' scope = both;
Note: Enabling the native full outer join implementation in the RDBMS is preferable to disabling the FULL_OUTER_JOIN_SUPPORTED in the RPD.
Oracle BI Applications under Oracle support only binary sorting. If you are running an Oracle client, do one of the following:
Set the NLS_SORT parameter to BINARY.
Choose a NLS_LANG setting that includes binary.
These settings are required for adequate performance from the dedicated Web client.
Make sure that cost-based optimization is enabled in the Oracle development, test, and production databases, and that statistics are kept up to date. Otherwise, the rule-based optimizer can be used.
Create foreign keys in the Oracle database, but configure Oracle to not enforce the foreign key relationship. The existence of foreign keys will allow Oracle to better optimize certain queries. By turning off enforcement, the database load should not be negatively affected.
Analyze application for occurrences of highly skewed data that is indexed. Create histogram statistics for these indexes to enable the optimizer to better perform queries.
To increase data throughput between Oracle BI Server and Oracle, change SDU and TDU settings in listener.ora. The default is 2 KB and can be increased to 8 KB.
On the server side, edit the listener.ora file. Under the particular SID_LIST entry, modify SID_DESC as follows:
SID_LIST_LISTENER = SID_LIST = SID_DESC = (SDU=16384)(TDU=16384) ORACLE_HOME = /.....) SID_NAME = SOLAP) ) )
Make sure the temporary tablespace has adequate space.
Set the number of log file groups to 4.
Set the size of each log file to 10 MB.
On the client side, edit the tnsnames.ora file. Modify the TNS alias by adding SDU= and TDU= as follows:
myhost_orcl.world= DESCRIPTION=(SDU=16384)(TDU=16384) ADDRESS = (PROTOCOL = TCP)(HOST=myhost)(PORT=1521)) CONNECT_DATA=(SID=ORCL))
This section contains guidelines that are specific to IBM DB2 UDB databases. Table 3-6 provides guidelines for parameter settings for DB2 relational database management system (RDBMS) usage. Use these guidelines as a starting point. You will need to make changes based on your specific database sizes, data shape, server size (CPU and memory), and type of storage. The database administrator should make changes to the settings based on performance monitoring and tuning considerations.
Table 3-6 Recommended DB2 Parameter Settings
Parameter | DB2 UDB V7 | DB2 UDB V8 and V9 | Notes |
---|---|---|---|
SHEAPTHRES |
400000 |
400000 |
|
ASLHEAPSZ |
15 |
15 |
|
RQRIOBLK |
65535 |
65535 |
|
QUERY_HEAP_SZ |
16384 |
16384 |
|
JAVA_HEAP_SZ |
2048 |
2048 |
|
MAXAGENTS |
400 |
400 |
|
NUM_INITAGENTS |
10 |
10 |
|
NUM_POOLAGENTS |
200 |
200 |
|
INTRA_PARALLEL |
YES |
YES |
|
FCM_NUM_BUFFERS |
12288 |
12288 |
|
SHEAPTHRES_SHR |
N/A |
=SHEAPTHRES |
|
DBHEAP |
16384 |
16384 |
|
CATALOGCACHE_SZ |
5558 |
5558 |
|
LOGBUFSZ |
2048 |
2048 |
|
UTIL_HEAP_SZ |
10000 |
10000 |
|
NUM_ESTORE_SEGS |
16 |
NIL |
Restore is not needed in DB2 V8 64-bit because the V7 limit of 1.75 GB addressable memory has been lifted. |
ESTORE_SEG_SZ |
65536 |
NIL |
|
LOCKLIST |
25000 |
25000 |
|
APP_CTL_HEAP_SZ |
5000 |
5000 |
|
SORTHEAP |
4000 |
4000 |
|
STMTHEAP |
40960 |
40960 |
|
APPLHEAPSZ |
2560 |
2560 |
|
PCKCACHESZ |
2560 |
2560 |
|
STAT_HEAP_SZ |
20000 |
20000 |
|
DLCHKTIME |
10000 |
10000 |
|
MAXLOCKS |
50 |
50 |
|
LOCKTIMEOUT |
1200 |
1200 |
|
MAXAPPLS |
500 |
500 |
|
AVG_APPLS |
10 |
10 |
|
MAXFILOP |
500 |
500 |
|
GROUPHEAP_RATIO |
N/A |
70 |
New in V8 |
APPGROUP_MEM_SZ |
N/A |
30000 |
New in V8 |
DATABASE_MEMORY |
N/A |
AUTOMATIC |
New in V8 |
Note:
To avoid fatal deadlocks during the ETL, make sure that you select the 'Session Level Retry on Deadlock' option in Informatica.
This section contains guidelines that are specific to IBM DB2 UDB zOS databases. The following requirements apply to IBM DB2 RDBMS usage for zOS and OS/390:
The Oracle BI Applications communicate with IBM DB2 UDB for z/OS and OS/390 (running on zSeries servers) through IBM DB2 Connect middleware.
The following editions of DB2 Connect are supported:
DB2 Connect Enterprise Edition (EE). This edition is installed on a mid-tier server such as an Informatica Server/Client, DAC, and Oracle Business Intelligence.
DB2 Connect Unlimited Edition (UE). This edition provides the functionality of DB2 Connect Enterprise Edition but is priced differently.
The ODBC driver for all connections must use the IBM DB2 ODBC Driver.
Make the appropriate connections using the DB2 Client Configuration Assistant.
Use the variable settings shown in Table 3-7.
Table 3-7 Variable Settings for IBM DB2 UDB zOS and OS/390 Databases
Parameter | Recommended Setting | Notes |
---|---|---|
IDTHTOIN |
1800 |
|
CDSSRDEF |
Any |
|
STARJOIN |
1 |
This setting indicates that star join is enabled. The one table with the largest cardinality is the fact table. However, if there is more than one table with this cardinality, star join is not enabled. |
This section contains guidelines that are specific to SQL Server databases, and contains the following topics:
Note:
The SQL Server database must be created with a collation sequence that supports binary sort order or case-sensitive dictionary sort order. Case-insensitive dictionary sort order is not supported. For example, for binary sort order with the U.S English character set, use the collation 'Latin1_General_BIN'. If you use the default collation setting of 'SQL_Latin1_General_CP1_CI_AS', the database is set to case-insensitive, which is not supported, and causes index creation failures.
Oracle BI Applications requires that SQL Server databases be created with the ANSI NULL option selected.
To set the ANSI NULL option
In a SQL Server 2000 environment, when loading Oracle BI Applications tables with international data, or loading more than one language, you need to modify the DB Library Options setting.
To modify the DB Library Options setting
From the Microsoft SQL Server program menu, select Client Network Utility.
Select the DB Library Options tab.
Clear the option Automatic ANSI to OEM.
Note:
SQL Server 2000 automatically tunes many of the server configuration options; therefore, an administrator is required to do little, if any, tuning. Although these configuration options can be modified, the general recommendation is that these options be left at their default values, allowing SQL Server to automatically tune itself based on run-time conditions.
If necessary, SQL Server components can be configured to optimize performance, as shown in Table 3-8.
Table 3-8 Recommended Variable Settings for SQL Server Databases
Parameter | Recommended Setting | Notes |
---|---|---|
Affinity mask |
0 |
|
Allow updates |
0 |
|
Awe enabled |
0 |
|
C2 audit mode |
0 |
|
Cost threshold for parallelism |
5 |
|
Cursor threshold |
–1 |
|
Default full-text language |
1033 |
|
Default language |
0 |
|
Fill factor |
95% |
For insert-intensive transactions, set Fill Factor between 90 and 95%. For better query performance, set Fill factor to 95 or even 100%. |
Index create memory |
1024 KB |
Default is 0. |
Lightweight pooling |
0 |
|
Locks |
0 |
|
Max degree of parallelism |
0 |
Default is 0. This turns off parallelism. Max degree of parallelism should be left at 0, which means use parallel plan generation. It should be set to 1 (use only 1 process) if you run multi threaded components (for example, several EIM threads). |
Max server memory |
2000 MB |
Default is 2147483647. |
Max text repl size |
65536 B |
|
Max worker threads |
100 |
Default is 255. |
Media retention |
0 |
|
Min memory per query |
1024 KB |
|
Min server memory |
500 MB |
Default is 0. |
Nested triggers |
1 |
|
Network packet size |
8192 B |
Default is 4096. |
Open objects |
0 |
|
Priority boost |
0 |
|
Query governor cost limit |
0 |
Modify to 60 only if CPU is high. |
Query wait |
–1 sec |
|
Recovery interval |
0 min |
|
Remote access |
1 |
|
Remote login timeout |
20 sec |
|
Remote proc trans |
0 |
|
Remote query timeout |
600 sec |
|
Scan for startup procs |
0 |
|
Set working set size |
0 |
|
Two-digit year cutoff |
2049 |
|
User connections |
0 |
|
User options |
0 |
SQL Server memory: Make sure adequate memory is available.
Transaction logs and TempDB: Reside on a separate disk from those used by database data.
Full load: Full Recovery model for the database.
Incremental (Refresh) load: Change from Full to Bulk-Logged Recovery model.
This section contains guidelines that are specific to Teradata databases, and contains the following topics:
Section 3.8.1, "Installation of Required JDBC Driver for Teradata Databases"
Section 3.8.2, "General Guidelines for Teradata Deployments"
The Data Warehouse Administration Console (DAC) requires JDBC drivers for database connectivity. You should only use JDBC drivers that are compatible with the supported databases. For information about supported databases, see System Requirements and Supported Platforms for Oracle Business Intelligence Applications. Since JDBC drivers show variations with different database versions, only drivers that are shipped with the database or downloaded from the database vendor site and are known to be certified for the given database version should be used. Currently, third-party JDBC drivers for the databases are not supported.
If you have a Unicode environment on a Teradata database, you must install the Teradata JDBC Driver for Teradata 12.0 or Teradata 13.0. This driver is required for all supported versions of Teradata databases, including versions earlier than version 12. The Teradata JDBC Driver for Teradata 12.0 and Teradata 13.0 is available at http://www.teradata.com/DownloadCenter, under the title Teradata JDBC Driver.
The following requirements apply to Teradata database usage:
If a deadlock issue is encountered, Oracle recommends that you use the 'Execute Serially' option for the specific groups within DAC. For Teradata, this is mandatory. When this option is selected for the task group, you must rebuild and run the Execution Plan that includes that task.
Install the Teradata Parallel Data Pump (TPump) Teradata Load utility on the machine on which the Informatica Server is installed.
When installing Informatica PowerCenter, make sure there are no spaces in the Informatica Server directory name or the directory path. The default directory contains spaces that you must remove manually.
When you install Oracle BI Applications in a Teradata environment, you must load the pre-built Informatica Repository file Oracle_BI_DW_Teradata.rep, which is located in ORACLE_HOME\biapps\dwrep\Informatica\Repository.
In the ODBC configuration for Teradata, set the session Mode to ANSI and DateFormat to AAA, before you create the schema. Tables must be created as case specific. If you do the ODBC configuration after the schema is created, tables might not be created as case specific.
To avoid overflows if the arithmetic calculation involves any decimal data type more than 18,3 (precision, scale), add the following static source system parameters:
$$Hint_Tera_Post_Cast = "as Decimal(18,3))"
$$Hint_Tera_Pre_Cast = "Cast("
For more information about setting source system parameters in DAC, see Section 4.19.2, "Setting DAC Source System Parameters."
Install the reswords.txt file in the server/bin directory for Informatica. When you are configuring the reswords.txt file, note the following points:
If any table name or column name contains a database reserved word, such as MONTH or YEAR, the session fails with database errors when the Informatica Integration Service executes SQL against the database. You can create and maintain a reserved words file, reswords.txt, in the server/bin directory. When the Integration Service initializes a session, it searches for reswords.txt. If the file exists, the Integration Service places quotes around matching reserved words when it executes SQL against the database.
Use the following rules and guidelines when working with reserved words:
The Integration Service searches the reserved words file when it generates SQL to connect to source, target, and lookup databases.
If you override the SQL for a source, target, or lookup, you must enclose any reserved word in quotes.
You might need to enable some databases, such as SQL Server, to use SQL-92 standards regarding quoted identifiers. Use connection environment SQL to issue the command.
For example, use the following command with SQL Server: SET QUOTED_IDENTIFIER ON Sample reswords.txt File.
To use a reserved words file, create a file named reswords.txt and place it in the server/bin directory. Create a section for each database that you need to store reserved words for. Add reserved words used in any table or column name. You do not need to store all reserved words for a database in this file. Database names and reserved words in reswords.txt are not case sensitive.
A sample reswords.txt file is listed below:
[Teradata] MONTH DATE INTERVAL
Note:
To avoid fatal deadlocks during ETL processes, make sure that you select the 'Session Level Retry on Deadlock' option in Informatica.
This sections contains the recommended best practises for maximizing performance in Teradata deployments, and contains the following topics:
Note:
The following best practices should be treated as customizations. You should follow standard customization methodology, such as copying the mapping to a custom directory. You should never make these changes directly in the default objects.
Once the tables have been created in the staging and target databases, you must run the supplied statistics collection. Failure to do so can affect ETL performance and possibly result in a spool space error (error number 2646).
DAC does the recollection of statistics as part of the ETL processes. However, DAC issues the collect statistics statement at the table level only (for example, collect statistics on w_org_d), and only for existing statistics.
Teradata code performance is highly dependent on the specific environment of each installation. A high number of occurrences of a single (or few) values in columns that are involved in joins, whether null or not, might cause skewing of data across Teradata AMPs. The impact of this is increased likelihood of exceeding the "per AMP" spool limit as well as increased CPU usage on one AMP while the other AMPs are much less utilized for the query experiencing the skew. This extends the processing time for this query, and negatively impacts other queries in the system that are competing for CPU resources on the AMP experiencing the skew.
Depending on the environment, the current code might redistribute the outer table on the join key, except when the inner table is very small; the Teradata Optimizer might choose to copy the inner table to all AMPs and not redistribute the outer table. Excessive nulls or other values in foreign keys will skew data in Teradata during join operations. If this happens, verify that statistics were defined and collected on the tables involved. If all the necessary statistics are defined and recently collected, it might be necessary to rewrite the SQL.
Many fact SIL mappings need to get the dimensional ROW_WID from ROW_ID/INTEGRATION_ID. For example, W_PER_RANK_FS.ACCNT_ID needs to be converted to ACCNT_WID before loading into the W_PER_RANK_F table. Since ACCT_ID is nullable, the join between W_PER_RANK_FS and W_ORG_D is defined as a LEFT OUTER JOIN.
However, the percentage of NULLs in ACCT_ID column can be as high as 50 percent or more depending on the data set. When redistributing the W_PER_RANK_FS according to ACCT_ID, all rows with ACCT_ID = NULL are put onto a single AMP.
Although a Teradata database usually has hundreds of gigabytes of spool space, the spool space is allocated across hundreds of AMPs. The spool space for each AMP is limited (for example, to two gigabytes).
When a large percentage of W_PER_RANK_FS is distributed to a single AMP, this can result in insufficient spool space being available. This happens when too much data is spooled onto one AMP, not because the spool space is too small.
To work with Teradata's mechanism of parallel processing and resolving the LEFT OUTER JOIN, SQL must be re-written.
As an example, refer to the following original SQL:
SELECT ... FROM W_PER_RANK_FS FS LEFT OUTER JOIN W_ORG_D ORG ON FS.ACCNT_ID = ORG.INTEGRATION_ID AND FS.DATASOURCE_NUM_ID = ORG.DATASOURCE_NUM_ID
The above SQL should be re-coded to convert the NULLs to some evenly distributed but non-matched values, as shown in the following SQL example:
SELECT ... FROM W_PER_RANK_FS FS LEFT OUTER JOIN (SELECT FS.INTEGRATION_ID, FS.DATASOURCE_NUM_ID, ORG.ROW_WID, ORG.GEO_WID FROM W_PER_RANK_FS FS, W_ORG_D ORG WHERE FS.ACCNT_ID = ORG.INTEGRATION_ID AND FS.DATASOURCE_NUM_ID = ORG.DATASOURCE_NUM_ID AND FS.ACCNT_ID IS NOT NULL) ORG ON FS.DATASOURCE_NUM_ID = ORG.DATASOURCE_NUM_I AND FS.INTEGRATION_ID = ORG.INTEGRATION_ID
The same SQL re-coding method can be used for other Source Qualifiers that run into spool space problems.
When there is a low number of distinct values, it is more efficient to use the GROUP BY phrase. Do not use the DISTINCT phrase, unless the number of distinct values is high.
If you do not use all of the preconfigured fields supplied, you can improve performance by stripping the extraneous fields from the mappings and tables.
This section explains the loaders that are available in Teradata, and how they are used in Oracle Business Intelligence Applications.
Teradata has three different types of Teradata loader processes, as follows:
Teradata Parallel Data Pump (Tpump) - for more information, see Section 3.8.3.5.1, "Tpump".
Fastload - for more information, see Section 3.8.3.5.2, "Fastload".
Mload - for more information, refer to the Teradata documentation.
Each loader process can be used in two different modes, as follows:
Staged Mode: The Informatica process does the following in this order:
Reads from the source data.
Creates a data file.
Invokes the loader process to load the table using the data file created.
Advantages: In the event of failures, you can recover using the Teradata recovery process.
Disadvantages: Staged mode is slower than Piped mode, and you need more disk space, as it can create large data files.
Piped Mode: The Informatica process reads from the source and simultaneously pipes that data to the loader to start loading the target table.
Advantages: Quicker than Staged mode, and you do not require large amounts of disk space because no data files are created.
Disadvantages: In the event of failures, you cannot recover using the Teradata recovery process (because tpump does row commits unlike fastload and mload).
TPump is a data loading utility that helps you maintain (update, delete, insert, and atomic upsert) the data in your Teradata database. TPump allows you to achieve near real-time data in your data warehouse.TPump uses standard Teradata SQL to achieve moderate to high data loading rates to the Teradata Database. Multiple sessions and multistatement requests are typically used to increase throughput.Unlike most load utilities, TPump uses row hash locks rather than table level locks. This allows you to run queries while TPump is running. This also means that TPump can be stopped instantaneously.
Tpump can be used in the following modes:
Tpump_Insert: Use to do inserts.
Tpump_Update: Use to do updates (this mode requires you to define the primary key in the Informatica target table definition).
Tpump_Upsert: Use to do update otherwise insert (this mode requires you to define the primary key in the Informatica target table definition).
Tpump_Delete: Use to do deletes (this mode requires you to define the primary key in the Informatica target table definition).
Informatica uses the actual target table name to generate the error table and log tables to be used as part of its control file generation. If you have two instances of Tpump loading into the same target table at the same time, then you need to modify the session to use a different error table and log table name.
The Tpump load process in piped mode is useful for incremental loads, and where the table is not empty. In the event of errors, restart the process and it starts re-loading from the last committed data.
Refer to Informatica documentation for information about configuring a session to use Teradata loaders.
The Fastload External Loader process is used on empty tables, such as loading staging tables and in initial loads where the tables are empty. When the Fastload process starts loading, it locks the target table, which means that processes (for example, lookups) cannot access that table. One solution to this problem is to specify dummy SQL for the look up overrides at the session level.
Tip:
If a session fails during a Fastload process, use SQL Assistant to run a simple SQL command (for example, count(*)), to determine whether the table is locked by a Fastload process.
If a table is locked (for example, for W_ORG_DS), use the following script to release the lock:
LOGON DATABASEALIAS/USER,PASSWORD BEGIN LOADING USER.W_ORG_DS ERRORFILES USER.ET_W_ORG_DS,USER.UV_W_ORG_DS; END LOADING;
If you save the above text in a file called test.ctl, you would run this process by entering the following command at a command prompt:
C:\fastload\test.ctl
Tip:
To create a load script for a table, edit the test.ctl script above to change the login information, and replace all occurrences of W_ORG_DS with the required target table name.
After a load process script runs successfully, you should be able to run the command 'select count(*)' on the target table. If you are not able release the lock, you might need to drop and re-create the table to remove the lock. If you do so, you must re-create the statistics.
Tip:
Fastload is typically used in piped mode to load staging tables and initial loads. In the event of errors, reload the entire data.