Optimize Disk Usage

Disk problems can occur when disk I/O activity is high. Common problems include disk fragmentation and chaining.

To avoid disk problems, especially when processing a large model:

Spread the workload as evenly as possible across multiple disk drives to reduce physical I/O contention by allowing parallel I/O operations. For example, if you have four physical disk drives, you could separate the operating system, database log files, CSE tables, and CSE indexes. See the table Recommended Configuration for Optimal Disk Performance.
If you have only a single drive, consider obtaining additional drives. The more disk drives you have, the less time processes spend waiting for disk I/O operations.
Strive to ensure that all disks are operating within their recommended maximum I/O rates.
If only certain disks exceed recommended maximum I/Os, move tablespace from the overworked disks to other disks.
Ensure that tables and indexes are assigned storage locations of adequate size. If sized correctly, all data will be in one contiguous extent rather than fragmented. Unfragmented data provide the most efficient access by eliminating the need for extra read/write head movement.
Plan for future expansion. If you need additional disks, consider adding several medium size disks rather than a few large disks to allow for better sharing of the workload.
Add memory for I/O buffering to reduce disk activity.
Adjust buffer caches as needed, either by adding more or reducing the current number.
Provide ample amounts of embedded free space.
Ensure that an adequate amount of system sort workspace is provided.
If available for your hardware platform, take advantage of Array disk technology.

Note: While it is possible to configure the CSE environment such that the database is located on a different physical machine, it is not a recommended configuration for performance reasons. The architecture of the normal execution of the CSE process is such that there are a very large number of individual messages passed between the CSE process and the database process. This leads to performance issues since any network latency between the two machines will be multiplied by the number of requests made. If the current configuration is absolutely necessary, then care should be taken to ensure that the communications infrastructure between the two machines is as quick as possible.