Publication Summary and Abstract
Ben Mitchinson, Tak-Shing Chan, Jon Chambers, Mark Humphries, Kevin Gurney and Tony Prescott (2008), BRAHMS: Novel middleware for integrated systems computation, Frontiers Neuroinformatics Conference Abstract: Neuroinformatics.
Computational modellers are becoming increasingly interested in building large, eclectic, biological models. These may integrate nervous system components at various levels of description, other biological components (e.g. muscles), non-biological components (e.g. statistical discriminators or control software) and, in embodied modelling, even hardware components, all potentially with different authors. There is a need for middleware to facilitate these integrated systems. BRAHMS (brahms.pbwiki.com), a modular execution framework, fills that need by defining a supervisor-process interface and an (extensible) set of process-process interfaces; authors can write to that interface, and processes will integrate when required. Additional benefits include: reuse (never code the same model twice); cross-user readability; system-level parallelisation on multi-core or multi-node environments; cross-language integration; data compression; data logging; performance analysis; run-stop-examine-continue execution. BRAHMS is also built around the nascent, and similarly general purpose, model markup language, SystemML (systemml.pbwiki.com). This will, in future, also facilitate: repeatability (same answers ten years from now); transparent automatic software distribution; interfacing with other SystemML tools.
Article via Digital Object Identifer (DOI) - article retrieved is probably subject to publisher's copyright