Recent Webcasts
Ivona Brandic, Data-Driven Techniques for Fault Tolerance in Edge Computing Systems - Theory
Fog Tutorial for Industrial IoT, Embedded Systems Week (ESWEEK 2021)
About Us
We develop foundational mechanisms for the energy-efficient resource allocation in edge infrastructure focusing on resource scarcity of the edge, which invalidates the most related resource allocation mechanisms intended for cloud. We define new mechanisms for approximative computing on the edge addressing the issues of incomplete data, fault tolerance and different computational models such as computation/data offloading, replication or handoff. Members of the HPC Research Group employ a wide range of state-of-the-art tools in their research including; artificial intelligence, multi-objective optimization, Monte Carlo simulation, time series analysis, control theory, fuzzy logic, formal methods, etc. This resulted in influential publications in several top-ranked venues such as SIGMETRICS, TPDS, FGCS, TCC, TSC, TNSM, CCGrid, or ICPE to name a few.
Research Activities
Our research topics include, but are not limited to:
- Trustworthy and sustainable Edge/Cloud systems
- Near real-time data analytics
- Virtualized HPC systems
- Massive-scale data analytics
- Energy efficient ultrascale distributed systems
- Cloud, Web & Workflow Quality of Service (QoS)
- The utilization of Blockchain for arbitrary resource allocation
Experimental testbed
Through different testbeds we build a novel distributed edge analytics framework and allow dynamic and self-adaptive placement of processing components across edge nodes. Different edge nodes are connected with Netgear 24-Port 10-Gigabit Switch.
Heterogeneous devices
We employ heterogeneous edge devices such as Raspberry Pi 3B+ in three clusters, Raspberry Pi 4B with attached 5MP cameras, Edge TPUs (USB accelerators), Up Squared AI Edge, Nvidia Jetson Nano, 5G router, 5G-enabled smartphone, Digital Power Meters.
Our Equipment
We focus on different applications and use cases ranging from smart buildings to video analytics in smart cities. We test system capabilities while respecting near real-time constraints on applications with higher demands (e.g., accuracy, latency).
Our Servers
For compute-intensive application tasks, we utilize our 12 in-house heterogeneous servers and GPU-dedicated server including RTX 5000, RTX 3090 and RTX 8000.
Ivona Brandić: Fog Tutorial for Industrial IoT
Data-driven techniques for fault tolerance in Edge computing systems - Theory
Embedded Systems Week (ESWEEK 2021)
Oct 8, 2021