Improving Latency in Mobile/Cloud Applications

Abstract

Smartphones are becoming comparable to desktop computers in terms of computational power, and offer diverse applications ranging from social media and gaming, to multimedia and banking. A particular class of mobile applications, mobile/cloud applications, are tightly coupled with the cloud. While executing on the mobile device, they communicate frequently with the cloud for crucial functionality.

Use of cloud-provided services is integral to the operation of mobile/cloud applications. And while the computational power of the cloud is seldom a performance concern, the network latency incurred when connecting a mobile device to the cloud can cause perceptible delays in the application. Users have a low tolerance for delays, so avoiding user-perceived delays is imperative to stop users from switching application providers.

This dissertation presents the Jovaku system, which aims to reduce communication latency between mobile devices and cloud services in a generic way, and by reusing existing infrastructure. Jovaku consists of a middle tier component designed to optimize mobile/cloud interactions and a Software Development Kit (sdk) that allows developers to leverage its capabilities.

The viability of the Jovaku system is substantiated through implementation of several modern mobile/cloud applications. Picster and Dapper both make use of Jovaku to reduce communication latency with their respective cloud services. We also perform an extensive experimental evaluation of Jovaku, revealing latency reduction by as much as 72% for certain mobile/cloud applications.