Check out the new USENIX Web site. next up previous
Next: Methodology Up: The Case for Compressed Previous: Adapting to Recent Behavior

Detailed Simulations

In this section, we describe the methodology and results of detailed simulations of compressed caching. We captured page image traces, recording the pages touched and their contents, for six varied UNIX programs, and used these to simulate compressed caching in detail.

(The code for our applications, tracing and filtering tools, and compressors and simulator are all available from our web site for detailed study and further research.)

Note that our traces do not contain references to executable code pages. We focus on data pages, because our main interest is in compressing in-memory data. As we will explain in Section 5, compressing code equally well is an extra complication but can certainly be done. Several techniques complementary to ours have been proposed for compressing code and the data from [RC96] indicate that references to code pages exhibit the same locality properties as references to data pages.


Scott F. Kaplan