A common requirement of the 4 million apps running on the world’s 2 billion smartphones is persisting structured data. Embedded databases such as SQLite are heavily used for this purpose, with a single typical Android smartphone averaging more than two SQLite queries per second. The fundamental challenges experienced by mobile apps using embedded databases - minimizing energy consumption, latency, and disk utilization - are familiar ground for database researchers. However, in spite of active research in the areas of smartphone query processing and embedded databases, the specific tradeoffs introduced by this new domain of pocket-scale data are far less well understood.
Key challenges in this space include the lack of publicly available data regarding smartphone database usage patterns in the real world, concrete high-level optimization targets, and tools and methodologies for reliably measuring database performance along axes relevant to smartphone apps. We propose infrastructure support and community-building efforts that will both improve existing research on embedded databases, and help to encourage new and innovative research in the area. This infrastructure support will take the form of real-world smartphone usage traces, a benchmarking suite for pocket-scale data, visualization tools, and instrumentation for mobile embedded databases.
Keywords: databases; smartphones; benchmarking.
The proposed infrastructure will be used by researchers from multiple academic and industrial institutions to support of new and existing research. Interest has already been expressed by researchers working on Adaptive Data Systems, Small Data Analytics, Gestural Query Processing, Data-Flow Analysis, Embedded Databases, Database Benchmarking, and others.
With 2 billion smartphones in the world, people are increasingly relying on smartphones to manage their lives. The proliferation of data-driven smartphone apps is driving a need to create more, better, faster, more user-friendly, and more power-aware techniques for managing their data. It is critical that we begin understand how smartphone apps interact with their data. Our proposal lays the groundwork for research on pocket-scale data management. We have interest from the Transaction Processing Council for our proposed benchmark, and even now several members of the database, systems, and programming language communities have expressed interest in the resources we propose to offer. In addition to supporting research in a critical area, this proposal will support one graduate student during the planning phase and up to two graduate students in later phases, resulting in between one and two PhD Theses.