Empirically assessing opportunities for prefetching and caching in mobile apps

Abstract

Network latency in mobile software has a large impact on user experience, with potentially severe economic consequences. Prefetching and caching have been shown effective in reducing the latencies in browser-based systems. However, those techniques cannot be directly applied to the emerging domain of mobile apps because of the differences in network interactions. Moreover, there is a lack of research on prefetching and caching techniques that may be suitable for the mobile app domain, and it is not clear whether such techniques can be effective or whether they are even feasible. This paper takes the first step toward answering these questions by conducting a comprehensive study to understand the characteristics of HTTP requests in over 1,000 popular Android apps. Our work focuses on the prefetchability of requests using static program analysis techniques and cacheability of resulting responses. We find that there is a substantial opportunity to leverage prefetching and caching in mobile apps, but that suitable techniques must take into account the nature of apps' network interactions and idiosyncrasies such as untrustworthy HTTP header information. Our observations provide guidelines for developers to utilize prefetching and caching schemes in app development, and motivate future research in this area.

Marauder

“…One class of systems employ program analysis techniques on app source code to determine both when and what to prefetch [9,31,37,61]. For example, Paloma [62] (described in more detail in §6.3) uses static analysis techniques and a local proxy to prefetch URLs one callback early. Similarly, APPx [9] uses static analysis to identify inter-request dependencies, and then (at a remote proxy) prefetches the dependent resources for each explicit request using online learning to fill in dynamic parameters.…”

“…Numerous studies have observed inefficiencies in the way that HTTP caches (both for web browsers and apps) operate [14,35,36,41,48,49,59,60,62]. These studies have shown that apps fall short of realizing the significant opportunities for caching [62] for a variety of reasons including lack of adherence to the HTTP caching specification [48,60], improperly set TTLs [35,36,49], and lack of support for aliased URLs that share the same content [26,30,35,36,41]. Our analysis of existing caching policies ( §3.3.1) mirrors these findings, and presents a fundamental tension in resolving the poor performance, i.e., ideal TTLs vary for a given resource.…”

“…• Caching [14,35,36,41,48,49,59,60,62], based on HTTP headers streamed from servers, enables mobile apps to store local copies of static resources and eliminate subsequent network fetches for those resources until they change. Although caching is widely used (74% of resources are cacheable in our experiments), existing policies are far from optimal, foregoing 52% of potential cache hits compared to policies that perfectly evict resources only when their content has changed.…”

Mobile-App Analysis and Instrumentation Techniques Reimagined with DECREE

“…Current mobile computing research has focused extensively on three threads: 1 static analysis techniques that analyze the apps' implementation artifacts statically to extract information of interest (e.g., security vulnerabilities [1], [2]); 2 instrumentation techniques that improve targeted aspects (e.g., performance [3]- [5]) of an app by directly modifying the app's implementation; and 3 auxiliary techniques that analyze external information associated with mobile apps to learn useful lessons (e.g., our recent work that assessed prefetching and caching opportunities [6]).…”

“…We have faced this gap in our prior research [2], [3], [5], [6]. The research community at large is also beginning to recognize this gap and the wasted opportunities it causes [1], [7]- [9].…”