Large-scale simulation of neuronal systems
2009 (English)Doctoral thesis, comprehensive summary (Other academic)
Biologically detailed computational models of large-scale neuronal networks have now become feasible due to the development of increasingly powerful massively parallel supercomputers. We report here about the methodology involved in simulation of very large neuronal networks. Using conductance-based multicompartmental model neurons based on Hodgkin-Huxley formalism, we simulate a neuronal network model of layers II/III of the neocortex. These simulations, the largest of this type ever performed, were made on the Blue Gene/L supercomputer and comprised up to 8 million neurons and 4 billion synapses. Such model sizes correspond to the cortex of a small mammal. After a series of optimization steps, performance measurements show linear scaling behavior both on the Blue Gene/L supercomputer and on a more conventional cluster computer. Results from the simulation of a model based on more abstract formalism, and of considerably larger size, also shows linear scaling behavior on both computer architectures.
Place, publisher, year, edition, pages
Stockholm: KTH , 2009. , xii, 65 p.
Trita-CSC-A, ISSN 1653-5723 ; 2009:06
IdentifiersURN: urn:nbn:se:kth:diva-10616ISBN: 978-91-7415-323-1OAI: oai:DiVA.org:kth-10616DiVA: diva2:221070
2009-06-09, Sal F2, KTH, Lindstedtsvägen 26, Stockholm, 10:00 (English)
Peck, Charles, Dr.
Ekeberg, Örjan, Docent
QC 201007222009-06-032009-06-032013-04-08Bibliographically approved
List of papers