IBM-BSC - Novel Technologies for Future Computing

Novel Technologies for Future Computing

Heterogeneous Cloud Ecosystem

This JSA focuses on resource management policies in a Kubernetes-based container cloud platform in which batch computing workloads can be easily deployed and efficiently managed. The roadmap includes continuation of the exploration on how to orchestrate heterogeneous workloads on the cloud, and on IBM OpenPOWER at the data center scale. The distributed workloads to be considered include high performance deep learning applications and material science exploration applications. The envisioned data centers are enabled with different types of compute accelerators, including Graphical Processing Units (“GPUs”). The activities are split into two lines of research: resource management and workload modelling. This project is led by Prof. David Carrera (BSC, Data Centric Computing Group Manager) and Dr. Alaa Youseff (IBM Research Yorktown, Container Cloud Platform).

FPGA-based innovations for Heterogeneous Computing

Heterogeneous computing systems are a key element to continued performance and energy-efficiency improvements in times of diminishing gains from silicon technology scaling. This JSA is focused on using reconfigurable computing based on FPGAs to improve performance and energy efficiency. The effort includes work on the programming model using OmpSs (porting Nanos/Picos to FPGAs) as well as the development and application-integration of relevant kernels for FPGA acceleration. This project is led by Prof. Miquel Moretó (BSC, Runtime-aware Architectures Activity Co-leader) and Prof. Xavier Martorell (BSC, Parallel Programming Models Group Manager) and Dr. Christoph Hagleitner (IBM Research Zurich, Cloud and Computing Infrastructure).