UNDERSTANDING HIV/AIDS PANDEMIC BY HIGH PERFORMANCE COMPUTING EPIDEMIOLOGY

摘要

This paper presents the application of distributed systems high performance cluster in modelling the spread of HIV in a given population. A population of a complex ecosystem such as a community of sexually related individuals has been modelled around distributed objects. An object modelled one individual and its communication with other objects represents social interaction between individuals. This research focuses on the heterosexual transmission of HIV and uses the probability of infection as determined by medical research work. The probability of infection is dependent on a number of factors; one) viral load of the infected person as well as two) the CD4+CCR5 receptor cell count of the person at risk and three) use of ARVs on mortality and transmission rate is also considered. The experiment was run on a cluster of 70 connected computers each capable of running hundreds of objects (individuals). These objects are implemented in Java in a distributed programming environment called CORBA that enable object across a network to communicate. The results obtained show patterns consistent with the epidemiological nature of the HIV as well as the behavioural properties of the people. The results have clearly demonstrated the ability for the high performance computer cluster to play out different scenarios and provide predictions and answers to "what if?" questions especially those of intervention programmes.