Check out the new USENIX Web site. next up previous
Next: Third experiment Up: Experimental results Previous: First experiment

Second experiment

 

  figure182


Figure 7: Ideal file arrival times vs. estimated file arrival times for Argo/UML (experiment 2).

In the second experiment, bundlings of the rt.jar subset were generated from 5 profiles collected from a single application, Argo/UML. Each profile exercised different application functionality. Then, arrival time data were collected for a single profile (which was a member of the profiles used to generate the bundlings).

Figure 7 shows the expected file arrival times vs. ideal file arrival times. As expected, the bundlings achieve file arrival times very close to the ideal file arrival times. Because of the regularity in class loading behavior in different runs of the same application, only a few bundles were generated, increasing the bundle size and compression ratio, and decreasing the cost associated with making requests to the server.

(Note that bundling 0.8-1000-500 is not shown because it is identical to 0.8-1000-200 for the profiles used.)



David Hovemeyer
Tue Feb 27 18:43:09 EST 2001