Datenschutzerklärung|Data Privacy
Impressum

19.11.2020
Martin Pagel

Di 01.12.2020 13:00 - 13:45 Uhr, Online: "Transfer Optimizations for Distributed Mega Data" (Niklas Semmler)

PhD thesis intro talk: "Transfer Optimizations for Distributed Mega Data"

Abstract:
In the Industrial Internet of Things (IIoT), data is generated at edge locations, locations that are connected via access networks and equipped with limited processing power. As the data production rate grows, mega-sized datasets of low-level sensor data emerge at the edge, that are too large to be transferred to the Cloud. For the success of the IIoT, data across multiple locations has to be jointly processed with limited footprint on the network.
In the thesis, we propose to combine existing data transfer paradigms with flexible aggregation to reduce transfer volume and delay. We focus on three transfer paradigms (a) synchronous messages, (b) query & replication and (c) dataflow that are currently used by a wide range of frameworks. In this talk, all three optimizations will be introduced on a high-level. Then we focus on transfer optimizations for the query & replication paradigm. We introduce online replication strategies that trigger proactive replication based on patterns found in access traces.
We evaluated heuristic, competitive and machine learning-based strategies on a real-world trace and show how these can be used to reduce the transferred data volume.

Bio:
Niklas Semmler is in the final stage of his PhD at SAP under the supervision of Anja Feldmann. He started his PhD at the TU Berlin, where he participated in the Software Campus programme. He received an MSc from KTH Stockholm and TU Berlin as part of the EIT Master School.
Previously he received an MSc from the University of Amsterdam (UvA).

Log-in information:
If you are interested in attending the online presentation, please contact web@bifold.berlin!