Technische Universität Darmstadt
Fachbereich Informatik
FG Datenbanken und Verteilte Systeme
Hochschulstr. 10 (Raum S2 02/E117)
64289 Darmstadt
Tel.: (0)6151-16-6232
Mail: hardock@dvs.tu-darmstadt.de

Flash-Memory Support for Virtual Spaces in Gaming and Sensor-Supported Environments

Humans perceive space through their senses and gather information from the environment. This information leads to knowledge that is then consciously or unconsciously used to determine their behavior. In today’s digital world our senses are supported and in many cases substituted by digital sensing devices. In all spheres of our life we use nowadays computing devices that act as sensors. These sensors produce data at a continuous rate. The huge volume of automatically and sensor generated data has lead to the need to handle “big data”.

All around us there are spaces that use sensors, or even more, which are defined by them. The current state in the evolution of sensor technique was possible thanks to increased computational power of processor units (while decreasing their form factor) and the invention of flash memory. Flash memory is a kind of persistent storage that is characterized through high performance, robustness, small form factor and low energy consumption – all of which are critical for the small, autonomous and usually wireless sensors. The optimal and integrated usage of the flash memory in the sensors, their networks and in large systems used for correlating sensor data is the main point of my project. The way how flash is used today is based on old concepts inherited from magnetic disk technology; it is based on providing the backwards compatibility with the traditional storage. The major drawbacks of this approach are: low performance (as compared to the raw flash potential) and lack of predictability, which becomes a critical issue for the spaces with real-time requirements using hundreds of sensors on flash.

The idea of this project is to move away from compatibility mode and use instead the raw flash to optimize data storage and processing on sensors and for sensor data. Our first results in this direction show the performance improvement of up to six times as compared to the current approach.

• Sensor-defined Spaces
• Flash Memory
• Big Data