|
Here is a typical database performance scenario: A key application gradually becomes less and less responsive over time. It's gotten to the point that the system, storage and database administrators are meeting to discuss the problem. Like the blind men describing the different parts of an elephant, each begins to analyze the problem from his own perspective.
The system administrator analyzes CPU, memory, paging and IO utilization, and (if trending data is available) confirms that utilization has been steadily increasing. The storage administrator reviews capacity and performance data, such as storage area network (SAN) switch port utilization, storage system cache parameters and physical disk utilization. It's discovered that the amount of data has grown significantly, and while there's plenty of SAN bandwidth, there's definitely a high rate of disk activity for the physical storage assigned to the database. The DBA begins to examine the problem from the application perspective and finds that queries and batch jobs are taking much longer to complete.
At this point, the next step is usually to go through several iterations of tuning. This often includes bringing in expertise from the vendors involved to make sure the system is optimized. Also, it isn't uncommon to begin to hear talk of the need for hardware upgrades. After several attempts, the conclusion is drawn that tuning options have been exhausted, and the only option is to upgrade.
However, there is another option: purge and archive the database. There are three classes of database archiving tools: ad hoc, native and third-party products.
An ad hoc process should only be considered for those rare instances when an archive will be only required infrequently and there isn't a need for long-term retention or frequent retrieval. Another consideration is the number of applications: More than one or two ad hoc database archiving processes will likely prove to be too difficult to manage.
Native tools are typically used for less complex, custom applications where data relationships are well understood within the organization. As long as data retrieval and retention requirements remain low, this approach is acceptable.
Third-party tools can make the archiving process more repeatable. They are particularly suited to large commercial applications, and where either regular access and/or long-term retention of archived data are required.
|
