Process of predicting and controlling the use of cache memory in a computer system

摘要

The cache memory space in a computer system is controlled on a dynamic basis by adjusting the low threshold which triggers the release of more cache free space and by adjusting the high threshold which ceases the release of free space. The low and high thresholds are predicted based on the number of allocations which are accomplished in response to I/O requests, and based on the number of blockages which occur when an allocation can not be accomplished. The predictions may be based on weighted values of different historical time periods, and the high and low thresholds may be made equal to one another. In this manner the performance degradation resulting from variations in workload caused by prior art fixed or static high and low thresholds is avoided. Instead only a predicted amount of cache memory space is freed and that amount of free space is more likely to accommodate the predicted output requests without releasing so much cache space that an unacceptable number of blockages occur.