RAM and Swap Disk Tips Donald Burleson

RAM and Swap Disk

Whatever the operating platform, the processing demands of the Oracle database must not exceed the real RAM memory of the server. All large servers share RAM resources through a Virtual Memory (VM) scheme. Oracle servers (Windows, UNIX, and OS-390) manage excessive RAM demands with a special swap disk, and virtual memory exploits the fact that not every executing task is constantly referencing its RAM memory region.

Since some RAM regions are accessed intermittently, vendors have developed paging algorithms that shift these memory pages to the swap disk when they are not in use.

The swap disk is a special disk area that provides for RAM sharing, primarily by storing page frames of inactive disk programs. The least-frequently-used (LRU) RAM pages are offloaded so that new applications can simultaneously share the same memory. After inactive RAM frames are paged-out to the disk, the operating system can utilize the freed memory for other active tasks.

If the inactive program later resumes execution, the RAM frames are paged-in from the swap disk.

This reloading of RAM pages from one memory area to another is called swapping. Swapping is very time-consuming and degrades performance of the application.

The swap disk allows simultaneous RAM usage greater than the real amount of RAM, but performance will be seriously compromised if the swap disk is used for active programs. The impaired performance is due to the much slower read times from the swap disk, compared to RAM. Disk access is measured in milliseconds, or millionths of a second, while RAM access is measured in nanoseconds, or billionths of a second.

Not all of this difference can be practically realized because of Oracle overhead, but most experts say that Oracle RAM is about 2,500 times faster than disk access.