Autonomic Systems and e-Business Platforms
The goal of our research group is to explore the future of computing by performing high-level research in today’s eBusiness applications that have to deal with critical IT challenges in areas such as Cognitive Computing, Big Data, Cloud Computing, High Performance Computing and Sustainable Computing. We organise our research in three main research lines:
Data-Driven Scientific Computing
The goal of this area is to design resource management strategies for big data applications, defining policies that enable distributed data stores to meet high-level performance goals. We focus on scientific applications, like those from life science domain, which data generation and accesses are both bound precision and performance. Currently, the main threads of our work are:
- Propose resource management strategies that are able to exploit novel hierarchical storage systems
- Hecuba: this project aims to design and develop strategies to facilitate programmers the efficient usage of data stores for big data applications. For example, define simple interfaces that are independent from the particular type of data system.
- Qbeast: this project consists on designing and implementing a distributed system to provide to multidimensional queries an efficient and reliable indexing mechanism.
Deep Learning and Artificial Intelligence with Supercomputing
The research group aims to carry out research that enables the interaction across many different disciplines from algorithms to infrastructure. The hope is, that cross pollination will accelerate progress towards Deep Learning, and Artificial Intelligence in general, using advanced Supercomputing technologies.
We are doing basic research in a wide range of Artificial Intelligence areas such as Deep Learning, Probabilistic Graphical Models and Reinforcement Learning, using Supercomputing platforms.
Research on the best practices for production systems based in Deep Learning are proposed. At this moment we are providing solutions to industry for document classification, image segmentation, text classification and distributed training.
In addition, we are collaborating in the organization of research events in the area and disseminating these concepts through the teaching of university courses and dissemination activities.
Energy-aware Computing and Virtualisation
The goal of this area is to develop management algorithms for virtualised Data Centres in a large-scale distributed ecosystem running heterogeneous workloads that optimize their operation with respect to energy and ecological efficiency.
The work in this area is grouped in the following main lines:
- Performance analysis of Virtual Machines and Containers in a virtualised Data Centre running HPC and Big Data workloads
- Models for the assessment and forecasting of energy and ecological efficiency in a virtualised Data Centre at different levels
- Policies for the optimization of the scheduling and placement of Virtual Machines and Containers in physical nodes considering the energy and ecological efficiency factors
- Policies for the selection of Data Centre for remote placement of Virtual Machines and Containers in a Data Centre ecosystem considering the energy and ecological efficiency factors
- Integration of the cooling and power supply subsystems in the energy management strategy of Data Centres
- Integration of renewable energy sources in the energy management strategy of Data Centres
Autonomic Team @ Marenostrum: