The controller uses two algorithms for flushing cache: demand-based and age-based. The controller uses a demand-based algorithm until the amount of cached data drops below the cache flush threshold. By default, a flush begins when 80 percent of the cache is in use.

In System Manager, you can set the “Start demand cache flushing” threshold to best support the type of I/O used in your environment. In an environment that is primarily write operations, you should set the “Start demand cache flushing” percentage high to increase the probability that any new write requests can be processed by cache without having to go to the disk. A high percentage setting limits the number of cache flushes so that more data remains in cache, which increases the chance of more cache hits.

In an environment where the I/O is erratic (with data bursts), you can use low cache flushing so that the system flushes cache frequently between data bursts. In a diverse I/O environment that processes a variety of loads, or when the type of loads are unknown, set the threshold at 50 percent as a good middle ground. Be aware that if you choose a start percentage lower than 80 percent, you might see decreased performance because data needed for a host read might not be available. Choosing a lower percentage also increases the number of disk writes necessary to maintain the cache level, which increases system overhead.

The age-based algorithm specifies the period of time during which write data can remain in the cache before it is eligible to be flushed to the disks. The controllers use the age-based algorithm until the cache flush threshold is reached. The default is 10 seconds, but this time period is counted only during periods of inactivity. You cannot modify the flush timing in System Manager; instead, you must use the Set Storage Array command in the command-line interface (CLI).