Deriving Streaming Graph Algorithms from Static Definitions

Increasing volumes of data and the desire for realtime query capability make the development of efficient streaming algorithms for data analytics valuable. Streaming graph algorithms that avoid unnecessary recomputation through clever application of data dependency analysis are often more complex to...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:2017 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW) S. 637 - 642
Hauptverfasser: Ediger, David, Fairbanks, James P.
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 01.05.2017
Schlagworte:
Algorithm design and analysis
Computational modeling
Database languages
graph algorithm building blocks (GABB)
Graph analysis
GraphBLAS
Linear algebra
Mathematical model
Measurement
Optimization
streaming graphs
triangle counting
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Increasing volumes of data and the desire for realtime query capability make the development of efficient streaming algorithms for data analytics valuable. Streaming graph algorithms that avoid unnecessary recomputation through clever application of data dependency analysis are often more complex to derive than their static counterparts. This paper discusses a method to derive algorithms for streaming graph analysis from static formulations Combining tuned graph algorithms building blocks with an appropriate functional language, a graph query planner should be able to correctly implement most static and streaming versions of an algorithm from a single mathematical formulation. We provide a detailed analysis for the case of updating triangle counts in a streaming graph using linear algebra and an experimental evaluation in Julia.
DOI:10.1109/IPDPSW.2017.146