Skip to main content
. Author manuscript; available in PMC: 2014 Apr 20.
Published in final edited form as: Proc USENIX Symp Oper Syst Des Implement (OSDI). 2012:61–75.

Figure 3.

Figure 3

Average elapsed time for swap microbenchmarks (lower is better). This benchmark allocates a buffer using malloc, touches each page in pseudorandom order, and reads the pages of the buffer in order to check correctness. The numbers above the bars indicate relative slowdown relative to Linux.