Date of Award

2013-01-01

Degree Name

Master of Science

Department

Computer Science

Advisor(s)

Eric Freudenthal

Abstract

Mobile devices such as smartphones and tablets are energy and memory limited, and implement graphical user interfaces that are intolerant of computational delays. Mobile device platforms supporting apps implemented in languages that require automatic memory man- agement, such as the Dalvik (Java) virtual machine within Google's Android, have become dominant. It is essential that automatic memory management avoid causing unacceptable interface delays while responsibly managing energy and memory resource usage. Dalvik's automatic memory management policies for heap growth and garbage collection scheduling utilize heuristics tuned to minimize memory footprint. These policies result in only marginally acceptable response times and garbage collection signicantly contributes to apps' CPU time and therefore energy consumption. The primary contributions of this research include a characterization of Dalvik's \base- line" automatic memory management policy, the development of a new \adaptive" policy, and an investigation of the performance of this policy. The investigation indicates that this adaptive policy consumes less CPU time and improves interactive performance at the cost of increasing memory footprint size by an acceptable amount.

Language

en

Provenance

Received from ProQuest

File Size

53 pages

File Format

application/pdf

Rights Holder

MD ABU JAHID

Share

COinS