Sponge: A searchable P2P mobile app store using DHTs

Abstract

In this paper we propose a novel searchable P2P content storage for storing mobile apps. In the growing need of storing and distributing huge number mobile apps across large number of users, it has been deemed necessary to think of a storage that leverages enormous amount of content space that common people can share these days. In our proposal, we leverage popular P2P stores, such as standard DHTs, for storing apps. One problem with DTHs is that they support flat names (denoted as keys) for content objects, which do not make content objects searchable by their different attributes. When users look for a certain app in an app store, they do not necessarily look by their unique names, instead by a set of attribute-value pairs (multi-attribute queries). We devise a technique, called predicate hashing, by which we hash these attribute-value pairs into hashed keys that are in turn used to locate corresponding apps from the underlying DHT. By doing this we effectively instrument flat named DHTs into reverse ifndexable content store. To this end, we build Sponge, a Kademlia based P2P app store, and do simulation based experiments to show that Sponge can retrieve content against multi-attribute queries.