Muskel is a parallel programming library providing users with structured parallel constructs (skeletons) that can be used to implement efficient parallel applications. Muskel applications run on networks/clusters of workstations equipped with Java (1.5 or greater). The skeletons are implemented exploiting macro data flow technology.

Currently, muskel supports the following pre-defined skeletons:

• farm modelling embarrassingly parallel computations
• pipe modelling computations arranged in stages
• seq wrapping sequential computations (pure functions) in such a way they can be used as parameters for other skeletons

Recently, another skeleton has been added:

• map modelling embarrassingly parallel data parallel computations

In addition, muskel allows users to define their own skeletons. Exploiting the macro data flow implementation, users can define new skeletons by providing the data flow graph implementing the skeleton. The graph may be provided using proper Java code or compiled using a GUI.

Two versions of code are available at the moment:

• the first one, is the original muskel, only providing pipe and farm skeletons, no possibility to extend the skeleton set, always compiling to normal form
• the second one is the new muskel, with the map skeleton, in addition to pipe and farm, supporting the possibility to create new skeletons by directly providing the macro data flow graph

All versions of muskel can be downloaded and used under the GPL license.

• download the original muskel
• download the last version of muskel
• downloadmultithread version of muskel

Both versions work perfectly on clusters and network of workstations supporting Java (1.6 is required for the new, last version, 1.5 is enough to run the original muskel) without firewalls blocking RMI accesses.

I collect there the code of some sample applications using muskel (last version). The purpose is to demonstrate the expressive power of the library and to provide some hints on how “classical” muskel applications can be written to novices.

• Sample pipeline application computing square of successors of integers. Shows usage of managers, input stream and output streams as well as the correct way to wrap sequential code.
  • Code of the application main
  • Code of Inc and Square classes (sequential code wrappers)
  • Sample input and output stream manager

muskel is being discussed (in several different ways) in the following papers:

• Marco Danelutto and Giorgio Zoppi, Behavioural skeletons meeting Services, in: Proc. of ICCS: Intl. Conference on Computational Science, Workshop on Practical Aspects of High-level Parallel Programming, pages 146-153, Springer Verlag, LNCS 5101, 2008

• Marco Aldinucci and Marco Danelutto, Securing skeletal systems with limited performance penalty: the muskel experience, in: Journal of Systems Architecture 2008 (doi:10.1016/j.sysarc.2008.02.008)

• Marco Aldinucci, Marco Danelutto and Peter Kilpatrick, Management in distributed systems: a semi-formal approach, in: Proc. of 13th Intl. Euro-Par 2007 Parallel Processing, pages 651–661, Springer Verlag LNCS 4641, 2007

• Marco Aldinucci, Marco Danelutto and Patrizio Dazzi, Muskel: an expandable skeleton environment (2007), in: Scalable Computing: Practice and Experience, 8:4(325-341)

• Marco Aldinucci and Marco Danelutto, Skeleton based parallel programming: functional and parallel semantic in a single shot (2007), in: Computer Languages, Systems and Structures, 33:3-4(179-192)

• Marco Aldinucci and Marco Danelutto, The cost of security in skeletal systems, in: Proc. of Intl. Euromicro PDP 2007: Parallel Distributed and network-based Processing, pages 213-220, IEEE, 2007

• Marco Aldinucci and Marco Danelutto, Algorithmic skeletons meeting grids (2006), in: Parallel Computing, 32:7(449-462)

• Marco Danelutto and Patrizio Dazzi, Joint structured/non structured parallelism exploitation through data flow, in: Proc. of ICCS: Intl. Conference on Computational Science, Workshop on Practical Aspects of High-level Parallel Programming, Springer Verlag, LNCS, 2006

• Marco Danelutto, Irregularity handling via structured parallel programming (2005), in: Intl. Journal of Computational Science and Engineering, 3-4

• Marco Danelutto, QoS in parallel programming through application managers, in: Proc. of Intl. Euromicro PDP: Parallel Distributed and network-based Processing, pages 282-289, IEEE, 2005

• Marco Danelutto, Adaptive Task Farm Implementation Strategies, in: Proc. of Intl. Euromicro PDP: Parallel Distributed and network-based Processing, pages 416-423, IEEE, 2004

• Marco Aldinucci, Marco Danelutto and Paolo Teti, An advanced environment supporting structured parallel programming in Java (2003), in: Future Generation Computer Systems, 19:5(611-626) (this is the paper about Lithium, the predecessor of muskel)

* Marco Danelutto, Dynamic Run Time Support for Skeletons, in: Proc. of Intl. PARCO 99: Parallel Computing, pages 460-467, Imperial College Press, 1999 (this is the paper introducing macro data flow implementation for skeleton based programming environments)