Bookshelf Home | Contents | Index | PDF |
Performance Tuning Guide > Tuning Siebel EIM for Performance > General Guidelines for Optimizing EIM > Recommended Sequence for Implementing EIM ProcessesThe following sequence is recommended for implementing EIM processes:
Before you start optimizing EIM processes, make sure there are no network problems or server performance problems that can affect the results. Siebel Expert Services recommends using at least 100 MB network segments and network-interface cards (NICs) to connect the Siebel server and Siebel database server. In addition, Siebel Expert Services recommends using a network switch or similar technology, rather than a hub, to maximize throughput. Optimizing the .IFB FileWhen you have finished coding and testing the .IFB file to meet your business requirements, the next step is to optimize the .IFB file. The selected parameters in each section of the .IFB file determine the focus of each EIM task. The following recommendations are provided for each section of the .IFB file:
Checking .IFB File OptimizationOne method to find out if the .IFB file is optimized is to check the status of the records being processed in the EIM tables. This indicates if there are tables or columns that are being processed unnecessarily. The following query can be used to check the status of records in an EIM table: select count(*), IF_ROW_STAT from If many rows have a status of PARTIALLY IMPORTED it is likely that further tuning can be done by excluding base tables and columns that are not necessary. For example, two tests were run to IMPORT 5000 accounts from EIM_ACCOUNT table. The first test included all of the base tables while the second test only focused on the four necessary tables by including the following line in the .IFB file: ONLY BASE TABLES = S_ORG_EXT, S_ADDR_ORG, S_ACCNT_POSTN, S_ORG_TYPE The first test took 89 minutes to import (excluding the Updating Primaries step), while the second test only took 2 minutes to import (excluding the Updating Primaries step). Separating EIM Processes by OperationWherever possible, divide the EIM batches into insert-only transactions and update-only transactions. For example, assume that you are loading 50,000 records into an EIM table as part of a weekly process. 10,000 records represent new data and 40,000 records represent updates to existing data. By default, EIM can determine which records are to be added and which records are to be updated in the base tables, however, EIM will need to perform additional processing (through SQL statements) to make these determinations. If you were able to divide the 50,000 records into different batch numbers based on the type of transaction, you could avoid this additional processing. In addition, the columns being processed as part of the update activity might be less than those for the insert activity (resulting in an additional performance increase). To illustrate this, the .IFBs in the preceding example can be coded with the following sections: |
Performance Tuning Guide |