MUSIC - MUlti-SImulation Coordinator

The INCF has engaged in the development of software that allows large scale neuron simulators to communicate during runtime. The software, known as MUSIC (Multi-Simulation Coordinator) is a project of the INCF Program on Multi-Scale Modeling of the Nervous System.

The INCF MUSIC project provides a standardized software interface for communication between parallel applications for large-scale modeling and simulation of neural networks.  

 

             multisimulation.png                  

The Multisimulation Coordinator (MUSIC) is a standard interface for run-time exchange of data among parallel applications in a cluster environment. It is designed for interconnecting large-scale neuronal network simulators with each other or with other tools. Data may consist of events, such as neuronal spikes, or graded continuous values, such as membrane voltages. A typical usage example is illustrated on the top, where three applications (A, B, and C) execute in parallel while exchanging data via MUSIC. The software interface promotes interoperability by allowing models written for different simulators to be simulated together in a larger system. It enables re-usability of models or tools by providing a standard interface. As data are distributed over a number of processors, it is non-trivial to coordinate data transfer so that it reaches the correct destination at the correct time. The above figure illustrates how the connection pattern between applications in the left part of the figure results in the inter-processor communication pattern in the right part.  The task for MUSIC is to relieve the applications of handling this complexity.

 

Current and future simulators can make use of MUSIC - compliant general purpose tools and participate in multi-simulations – a possibility that is particularly attractive when:

  • Different parts of a complex nervous system model are optimally implemented in different simulators, and need to communicate with each other.
  • Post-processing of generated data is needed, where the amounts of data are too large for intermediate storage, and requires the simulator to pass the data directly to the post-processing module.

In addition, such a standard interface enables straight-forward independent third-party development and community sharing of interoperable software tools for parallel processing.

In the design of the standard interface, care has also been taken to facilitate easy adaptation of existing simulators.

The MUSIC library was released in March 2009. Prior to its first release, INCF actively solicited comments from the community on the proposed design and prototype specifications.

   INCF MUSIC
  • A standard for neural network simulator interoperability
  • Readily available open-source C/C++ API library for straight-forward adoption in new and existing tools
  • Patchbay port metaphor gives an intuitive abstraction of the underlying MPI-based communication.
  • Each simulator will be able to use its own semantics while MUSIC provides the means for the other components to access exposed data
  • Designed for easy and straight-forward adaptation in new and existing software
  • Designed to meet the requirements of large-scale, high-performance simulation environments




  
MUSIC was initiated following a recommendation from the INCF Workshop on Large-Scale Modeling. It was developed by the Computational Biology & NeuroComputing group (CBN) at The Royal Institute of Technology, Sweden, under the leadership of Örjan Ekeberg and Mikael Djurfeldt, in a collaborative partnership with the INCF.

 

Poster presented at the SfN2008 meeting in Washington DC, November 16-19, illustrating how MUSIC interconnects to NEST and MOOSE (GENESIS).

MUSIC poster from Neuroscience2008 [pdf]

 

For more information and download of the library visit MUSIC on INCF Software Center


INCF Secretariat contact: Pontus Holm, pontus.holm@incf.org

Document Actions
Send this Print this Add to favorites