SATURN: Spatio-temporal cloud store for big-data applications
- Title: SATURN: Spatio-temporal cloud store for big-data applications
- Financing: FCT
- Coordination: Paolo Romano
- Page: http://www.inesc-id.pt/projects.php?grp=II11&Ano=10&prj=II02062
- Duration: 3
- Start: 01/07/2016
- Description:The Saturn platform will make extensive use of four emerging hardware technologies, namely GPUs (and other specialized coprocessors), Transactional Memory, RDMA and Non-volatile Memory, and employ them, for the first time, in a synergetic fashion to achieve unparalleled scalability and efficiently levels. This ecosystem of emerging hardware technologies will be used pervasively across the software stack of the Saturn data platform, in order to accelerate crucial mechanisms like concurrency control, networking, data durability. Besides leveraging on emerging hardware technologies, Saturn will expose an innovative programming model, specifically designed to overcome the limitations of state of the art cloud analytics frameworks. Efficiently exploiting large-scale distributed platforms, such as the cloud, for big-data processing requires carefully balancing processing and communication, as well as coping with issues like network partitions, asynchrony and failure of individual components.