Modeling GRID Based Computer Services for Public Health Applications

Recently, there has been a growth in the examination of the potential for GRID-based computing solutions for large-scale public health services IT projects. Potential projects include the aggregation and dissemination of public health surveillance data, coupled with automation of reporting tasks between healthcare and public health entities. There are few existing public health system implementations that can be used to estimate the scale and complexity of services required to support large-scale public health GRID-based computing solutions. In the absence of such information, alternative techniques must be considered to provide a high degree of certainty that proposed system solutions are tenable. Large public health IT projects require reliable risk estimation to achieve eventual compliance with the desired operational performance, before system development commences. Our project provides an agent-based simulation model of a public health GRID system as an aid to system design verification and performance prediction, using the National Cancer Institute's cancer Biomedical Informatics Grid (caBIG™) for reference. The model assists in verifying the system architecture and estimating future production performance by capacity planning, under a variety of considered failure domains. The model was developed to simulate the service components within a public health GRID system encapsulating biomedical, radiation, chemical and environmental sensors with attached wireless networks and associated data storage services. Additionally the model incorporates a set of GRID based data retrieval and analysis applications. This agent-based model considers the number of concurrent GRID services, variable data flow volumes, simultaneous load considerations and network bandwidth constraints, in order to provide a prediction of transaction throughput. Future models will include finer grain simulation of GRID services with the ability to manipulate service availability patterns. The eventual complete model will enable potential public health GRID projects to be modeled over variations in topology and load while simulating various network and service interruptions.