Welcome to Ubispark Project
What is Ubispark?
Ubispark is a research project with the aim to enable forming local computing clusters of smartphones and other smart devices for the purpose of large-scale data processing.
How?
Since large-scale data processing relies on distributed computing, by reducing individual task size in a computing job we can run these tasks using devices with limited memory, such as smartphones and smart TVs.
Easy to Use
Ubispark will be implemented as a regular Android app that users can download and contribute their device's idle computing power to science. The app can be set to contribute only when on WiFi and while charging, for example.
Motivation
To motivate distributed computing on smart devices, we have investigated the computational power and energy consumption of smart devices. According to preliminary studies:
- Current smartphones are faster than supercomputers of the past
- Smartphone computational power is growing exponentially
As we see on the very right, IoT devices, even smart refrigerators are catching up with supercomputers of the past.
Ubispark already runs on smartphones and smart TVs
- Using an Android app allows us to run the Ubispark on a TV, tablet or a smartphone
- We can measure the energy consumption of a smart TV using a wall plug monitor
- To measure energy compution of smartphones, we use a hardware power instead of the battery to power phones
- Currently Ubispark supports a limited set of tasks
Preliminary results
We have studied the performance and energy efficiency of smartphones compared to servers in the Ubispark project:
- 9 Samsung Galaxy S4 smartphones matches server performance
- Allowing 1h of battery life of 9 Samsung S4 devices to be used for computing is equivalent to 1h of computing time on a dedicated server.
On the right, we can see that the computation energy on a cluster of smartphones is much lower than that of a server machine for the same task.
On the left, the computation time for the same task for a cluster of smartphones is higher than that of the server, and after a certain limit, inter-device communication becomes a bottleneck. In this case, it makes sense to break the task down to multiple clusters, or use multiple master devices.
- 9 Samsung Galaxy S4 smartphones equal only 7.2% of the electricity cost of the same task when run on a HP Proliant server
Research Team
The Ubispark Research team includes Dr. Eemil Lagerspetz, Ella Peltonen, Jarno Leppänen, Dr. Petteri Nurmi, and Prof. Sasu Tarkoma.
Most of us are also involved in The Carat Project.
Development Team
Ubispark is developed by Dr. Eemil Lagerspetz, and Jarno Leppänen. New contributors are welcome.
For more information contact Dr. Eemil Lagerspetz at cs helsinki fi.
See also: Eemil Lagerspetz at Department of Computer Science, University of Helsinki