Home | Previous Page | Next Page   Tuning Configuration for Resource Usage > Tuning Background-Process Parameters >

Managing the Logical Log

Managing the logical log requires you to decide the size of the set of logical-log files that make up the logical log on each coserver, and to decide how many logical-log files to create.

Estimating the Size of the Logical Log

To estimate the size of the entire logical log in kilobytes, use the following formula:

LOGSIZE = ((connections * maxrows) * rowsize) / 1024

For connections, substitute the number of connections for all network types that the sqlhosts file or by one or more NETTYPE parameters specify. For maxrows, substitute the largest number of rows to be updated in a single transaction. For rowsize, substitute the average row size in a transaction. Use the maxrows and rowsize factors to approximate the size of the longest transaction that might occur except as the result of an error.

Tune the size of the logical log to accommodate expected long transactions and divide the total size of the logical log on each coserver into appropriately sized logical-log files, as described in the following section.

For information about long transactions and using the configuration parameters that specify their maximum size, see Managing Long Transactions in the Logical Log.

Specifying the Size and Number of Logical Log Files

After you estimate the total size of the logical log on each coserver, you decide whether either the number of logical log files on each coserver or the size of each log file is more important to provide optimal performance as well as fast recovery.

The LOGSIZE configuration parameter specifies the size of each logical log file on each coserver. The LOGFILES configuration parameter specifies the number of logical log files in the logical log on each coserver.

Consider the following factors:

For additional information about the relation of logical-log file size and number to fast recovery, see Specifying Logical Recovery Configuration Parameters.

Managing Long Transactions in the Logical Log

If a long transaction cannot be contained in the set of logical-log files, the transaction is aborted and rolled back. During the long transaction's roll-back process, the database server may not be available to execute other transactions.

Two configuration parameters, LTXWHM and LTXEWHM, determine how the database server detects and manages long transactions:

The logical log can be configured to expand dynamically to prevent a long transaction from spanning all available log space without completing, which would hang the coserver. The value of the DYNAMIC_LOGS configuration parameter determines whether the logical log expands dynamically and how much manual control the administrator has over that expansion.

Tip:
The LOGSIZE, LTXHWM, LTXEHWM, and DYNAMIC_LOGS configuration parameters can be tuned with the onutil SET command while the server is running.
Home | [ Top of Page | Previous Page | Next Page | Contents | Index ]