Efficient Pre-Fetch and Pre-Release Based Buffer Cache Management for Web Applications

Springer-Verlag Lecture Notes in Computer Science, vol. 3043, pp. 360-369, May 2004 (SCI)

Younghun Ko, Jaehyoun Kim, and Hyunseung Choo

Abstract

Various block replacement schemes have been proposed so far to enhance hit ratios by maintaining blocks that will frequently be re-referenced at buffer cache in the future. Especially the LRU block replacement scheme is widely used due to its simplicity and good adaptability to the changes on the working set. However, LRU has a severe drawback that exhibits itself in performance degradation in access patterns with weak locality such as sequential and looping reference. In this paper, we propose an improved LRU buffer cache management scheme using pre-fetching and pre-releasing based on spatial locality. It is as simple as the LRU scheme and retains its characteristics. Trace-driven simulation experiments show that the hit ratios are improved up to about 24% (with the average of 18%) compared to the LRU scheme for the workloads many engineers use.

 

Keyword

 

 

View Full Text