Dynamic cache sizing on Unix

On Unix, the database server uses swap space and memory to manage the cache size. The swap space is a system-wide resource on most Unix operating systems, but not on all. In this section, the sum of memory and swap space is called the system resources. See your operating system documentation for details.

On startup, the database allocates the specified maximum cache size from the system resources. It loads some of this into memory (the initial cache size) and keeps the remainder as swap space.

The total amount of system resources used by the database server is constant until the database server shuts down, but the proportion loaded into memory changes. Each minute, the database server evaluates cache and operating statistics. If the database server is busy and demanding of memory, it may move cache pages from swap space into memory. If the other processes in the system require memory, the database server may move cache pages out from memory to swap space.

Initial cache size

By default, the initial cache size is assigned using an heuristic based on the available system resources. The initial cache size is always less than 1.1 times the total database size.

If the initial cache size is greater than 3/4 of the available system resources, the database server exits with a Not Enough Memory error.

You can change the initial cache size using the -c option. See -c dbeng12/dbsrv12 server option.

Maximum cache size

The maximum cache must be less than the available system resources on the computer. By default, the maximum cache size is assigned using an heuristic based on the available system resources and the total physical memory on the computer. The cache size never exceeds the specified or implicit maximum cache size, or the sum of the sizes of all open database and temporary files plus the size of the main heap.

If you specify a maximum cache size greater than the available system resources, the database server exits with a Not Enough Memory error. If you specify a maximum cache size greater than the available memory, the database server warns of performance degradation, but does not exit.

The database server allocates all the maximum cache size from the system resources, and does not relinquish it until the database server exits. You should be sure that you choose a maximum cache size that gives good SQL Anywhere performance while leaving space for other applications. The formula for the default maximum cache size is an heuristic that attempts to achieve this balance. You only need to tune the value if the default value is not appropriate on your system.

You can use the -ch server option to set the maximum cache size, and limit automatic cache growth. For more information, see -ch dbeng12/dbsrv12 server option.

Minimum cache size

Discuss this page in DocCommentXchange.