SpiNNaker network experiment library.


License
GPL-2.0
Install
pip install network_tester==1.0.2

Documentation

SpiNNaker Network Tester

Latest Version Documentation Build Status Coverage Status

'Network Tester' is a library designed to enable experimenters to quickly and easily describe and run experiments on SpiNNaker's interconnection network. In particular, network tester is designed to make recreating traffic loads similar to typical neural software straight-forward. Such network loads feature a fixed set of cores with a fixed set of multicast flows of SpiNNaker packets between them.

The following is a (non-exhaustive) list of the kinds of experiments which can be performed with ‘Network Tester’:

  • Determining how a network copes with different rates and patterns of packet generation. For example to determining the maximum speed at which a particular neural simulation may run on SpiNNaker without dropping packets.
  • Determining the effectiveness of place and route algorithms by finding ‘hot-spots’ in the network.
  • Characterising the behaviour of the network in the presence of locally and globally synchronised bursting traffic.

Installation

The latest stable version of network tester may be installed from PyPI using:

$ pip install network_tester

Documentation

The latest documentation can be read on ReadTheDocs.

Development

For development information see DEVELOP.md.