Who slowed down my app? a measurement study of response times of smartphone apps in India

Abstract

India has seen a rapid rise in the adoption of smart phones, with a majority of Indians accessing the Internet through smart phones. Smart phones have become one of the most important mediums of delivering essential services such as news, entertainment, and even payment services in India. However, it is not clearly understood how the quality of experience of individual users varies, especially considering the huge diversity of the Smartphone models, network conditions, and regions from which they are used. We quantify the Quality of Experience of using the android apps using response time i.e., the time needed to reflect UI changes corresponding to a user's action. This can, for example, the time needed to reflect an item has been added to cart for Add product to cart action on Amazon app. In this work, we design a tool called EvalApp which uses automation to record the response times of a total of 30 actions for 12 apps popular in India. We then crowd source this desktop app to a total of 41 users working from home from across north and central India and perform a causal analysis of the factors that affect the response times of actions. We _nd that in most cases, the response time is very strongly correlated with the ping round-trip time to the nearest Google server, thus indicating that the network is the most common bottleneck. We further identify that the distance from a major city is the other factor that strongly affects c latency. This is counter-intuitive in view of recent works on the mobile web in the US, which indentified Smartphone hardware as the primary bottleneck. We also do a few controlled experiments to observe that having a higher app version does not always ensure better response times. Our observations are likely to help designers of apps to better identify techniques of improving the Quality of Experience for users.