Datenschutzerklärung|Data Privacy
Impressum

30.11.2016
K. Forster

29. 11. 2016, Paper "Gilbert: Declarative Sparse Linear Algebra on Massively Parallel Dataflow Systems" accepted at BTW 2017

Till Rohrmann, Sebastian Schelter, Tilmann Rabl, Volker Markl, Gilbert: Declarative Sparse Linear Algebra on Massively Parallel Dataflow Systems .

... In this paper, we propose Gilbert, a distributed sparse linear algebra system, to decrease the imminent lack of analytic capacities. Gilbert offers aMATLAB○R -like programming language for linear algebra programs, which are automatically executed in parallel. Transparent parallelization is achieved by compiling the linear algebra operations first into an intermediate representation. This language-independent representation enables high-level algebraic optimizations. Different optimization strategies are evaluated and the best one is chosen by a cost-based optimizer. The optimized intermediate representation is then transformed into a suitable format for parallel execution. Gilbert generates execution plans for Apache Spark○R and Apache Flink○R , two massively parallel dataflow systems. Distributed matrices are represented by square blocks to guarantee a well-balanced trade-off between data parallelism and data granularity. ...

Die 17. Fachtagung „Datenbanksysteme für Business, Technologie und Web“, BTW, der Gesellschaft für Informatik (GI) findet vom 6. bis 10. März 2017 an der Universität Stuttgart statt.