Journal: Frontiers in Neuroinformatics

Article Title: NeuroManager: a workflow analysis based simulation management engine for computational neuroscience

doi: 10.3389/fninf.2015.00024

Figure Lengend Snippet: The SimMachine Class Hierarchy . A SimMachine is the combination of three inheritance lines of objects. In this example a first ancestral line (1) is composed of classes that provide basic communications; file transfers between host and remote; ability to do MATLAB compilations of Simulator files; basic job submission facilities; and job submission specific to a SGE Cluster. The second line (2) supplies knowledge of how to build and host Simulators. Line 1 and 2 produce a machine class that can host and run Simulators on an SGE cluster. Adding the third line (3) provides location information for the NEURON software installed on a specific cluster. Together they give rise to a class MachineMySGECluster01 which corresponds to an actual computational resource. Further modifications result in an object that executes in a specific queue in a HPC resource. Classes above the dashed line are supplied with the Core code and those below are user-supplied. Arrows point to the superclass. Full diagram tree in User Guide.

Article Snippet: The overhead of NeuroManager is variable; however, the largest bottleneck is the MATLAB compilation.

Techniques: Software

Journal: Frontiers in Neuroinformatics

Article Title: NeuroManager: a workflow analysis based simulation management engine for computational neuroscience

doi: 10.3389/fninf.2015.00024

Figure Lengend Snippet: NeuroManager setup . (A) Define at least one SimMachine. This example defines MachineMyMach1 as a sub-class of MachineGenericUNIX , with corresponding machine type of MACHINE_MYMACH1. (B) Define a Simulator class that specifies a particular Simulator/Free Parameters setup. In this example, it is a simulator called SimMySimulator that is MATLAB only. This class has a corresponding SimType of SIM_MYSIMULATOR. (C) Write a session script that constructs NeuroManager. Use the SimMachines defined in A and specify the number of Simulators and work directory for each SimMachine, construct the MachineSet with a specific SimType SIM_MYSIMULATOR defined in (B) . Finally, run the simulations defined by the Input Parameter Vectors in “ myParameterSets.txt.”

Article Snippet: The overhead of NeuroManager is variable; however, the largest bottleneck is the MATLAB compilation.

Techniques: Construct