AN EFFICIENT METHOD TO REDUCE STARTUP TIME OF APPLICATIONS

Abstract

An application usually takes certain time for its loading during which the executable file and the other related files are brought to the main memory and necessary initializations routines are run. However, it has commonly been observed that in an actual use of an application by a general user, several of the loaded components are not essentially used during the lifetime of its execution. This results only in an increased space taken by them in the memory as well as an increased burden during their loading into the memory in terms of CPU cycles and memory wasted and time taken to load and initialize them, thus increasing the overall startup time of the software. In this Thesis, we propose a methodology by which we track the user’s usage pattern, and by allocating certain weights to some specific static and dynamic parameters of the software and its usage style, we selectively determine the potential components that are most likely to be used in the subsequent execution of that software and load only those necessary components at next startup. Thus, each time the user runs the software, he is presented with the set of components that he is most likely to use in that specific execution. We have done a simulation based on Monte Carlo approach and we have observed a reduction of around 40-50% in software load time and an approximately 70-80% Hit Rate for the components requested by user against the actually loaded components. We then developed a test application and compared its startup times and memory requirements at startup, and found around 59% reduction in application startup time and 16% reduction in start time memory requirements when our proposed method is used.