You are here: Home / Dissemination / ABS Primesieve Case Study

ABS Primesieve Case Study

We completed our research for the parallel and distributed algorithm of the Prime Sieve of Eratosthenes with the journal paper entitled “A Design Pattern for Optimizations in Data Intensive applications using ABS and JAVA 8″ to Concurrency and Computation: Practice and Experience.

Throughout this study we investigated several possibilities of implementing the Prime Sieve of Eratosthenes algorithm for multi-core and distributed systems in cooperation with the ENVISAGE EU Project (http://www.envisage-project.eu/). The main objectives and scientific impacts in our research were:

  • Observing the drawbacks of the ABS backends (Erlang and Java) when implementing a pipeline version of the algorithm with a significant performance penalty when each prime number represents a different entity. This resulted in the development of the ABS-API and the current research of a new JAVA backend.
  • Observing the drawbacks of representing a data partition as an entity with its own address space even when several partitions were on the same physical machine. This resulted in the development of ABS-API-Remote with actors aware of their location, a therefore able to share the address space and significantly improve memory management.  We present a comparison between Actors that share the same message queue while on the same VM and Actors that do not.

 

 

  • Generalizing the location-aware optimization into a programming design pattern and a data-oriented abstraction for high-performance applications.

The results were disseminated in the following papers:

Serbanescu, Vlad, et al. “Towards Type-Based Optimizations in Distributed Applications Using ABS and JAVA 8.” Adaptive Resource Management and Scheduling for Cloud Computing. Springer International Publishing, 2014. 103-112.

Serbanescu, Vlad, et al. “A Design Pattern for Optimizations in Data Intensive applications using ABS and JAVA 8″ . Concurrency and Computation: Practice and Experience. (to appear)

Azadbakht, Keyvan, et al. “High Performance Computing Applications using Parallel Data Processing Units”. Proceedings of the 6th IPM International Conference on 
Fundamentals of Software Engineering (to appear)