Personal tools
You are here: Home Science & Projects Deisa Extreme Computing Initiative Projects 2010 - 2011 Scaling of cortex simulations on large cluster computers

Scaling of cortex simulations on large cluster computers

Project BRAINCOR
Research Area Bio Sciences
Principal Investigator(s) Prof. Anders Lansner
Institution(s)
  • KTH, Sweden
  • Forschungszentrum J├╝lich, Germany

Abstract

Even the honey bee brain has on the order of a million neurons communicating via a thousand times as many synapses and it has a very intricate structure. The human brain is about 10000 times larger and much more complex. The size and complexity of the brain simulation models are severely limited by the restrictions imposed by the capacity of personal and small-sized parallel computers. This proposal concerns running brain simulations on a very large Blue Gene/P cluster supercomputer.

Very large-scale simulations are necessary in order to better understand the neural dynamics and information processing going on in the brain, since these are to a large extent globally coordinated phenomena. Such simulations will also help understand and diagnose diseases and dysfunctions of the brain and give knowledge important for drug development and other therapy.

Neural network simulations have a potential to scale very well on cluster architectures since most computations are local and long-range communication is event based. Our goal is to establish a workflow for very large scale to full-scale brain simulations. We aim to simulate mouse cortex sized networks (20 million neurons, 100 billion synapses) comprising biophysically detailed components as well as a realistic cortical architecture on 64K cores of the Blue Gene/P. These brain simulations will be among the largest ever done. They will feature more realism in terms of components and network structure than other simulation models and they will perform as an associative memory, which is an important brain function. European research in the area of brain simulation and neuromorphic engineering is internationally leading and our proposal is well connected to current European funded activities. Moreover, plans are currently in place for proposing an EC flagship project on brain science including simulation and it is therefore important to initiate these activities now.

Document Actions