logo SBA

ETD

Archivio digitale delle tesi discusse presso l’Università di Pisa

Tesi etd-04172021-170725


Tipo di tesi
Tesi di dottorato di ricerca
Autore
RINALDI, LUCA
URN
etd-04172021-170725
Titolo
Extending the Actor Model with Parallel Patterns: a new model for multi-/many-core platforms
Settore scientifico disciplinare
INF/01
Corso di studi
INFORMATICA
Relatori
tutor Prof. Torquati, Massimo
correlatore Prof. Danelutto, Marco
Parole chiave
  • actor model
  • data stream processing
  • multi-cores
  • parallel patterns
  • parallel programming
  • programming model
Data inizio appello
26/04/2021
Consultabilità
Completa
Riassunto
Nowadays, parallel processing is mandatory to improve application performances at all levels of the software stack. However, dealing with concurrency and parallelism is challenging and error-prone, although multiple parallel programming models and frameworks are proposed over the years. In this thesis, we consider two radically different approaches to parallel programming, i.e., the Actor Model and the Parallel Patterns, and attempts to combine them into a new heterogeneous model. The Actor-based programming model is a well-established and high flexible concurrent model, in which Actors can be spawned and connected dynamically. Actor programmers often pay such great flexibility with additional effort in identifying and solve performance issues and sub-optimal communication patterns. On the other side, the Parallel Patterns offers ready-to-use solutions to recurrent programming problems (e.g., pipeline, map-reduce). However, the price to pay is reduced flexibility and additional constraints during the software development process. The programmer needs to design the parallel application(s) as a composition of Parallel Patterns. We analyzed the limitations of the Actor Model on multi/many-cores and investigated multiple approaches to achieve fruitful cooperation of the two models and maintaining, at the same time, the core features and advantages of the two approaches.
File