S.V. Ivanov, A.V. Kalyuzhnaya, S.S. Kosukhin
A promising direction in the development of decision support systems (DSS), is a building of systems based on the processing of the incoming data. The data sources can include measurement systems, data aggregation systems (including historical archives) and computational models. Herewith the result of computations is not strongly defined in the frame of traditional algorithmic approach but can be obtained by different ways (scenarios) which option depends on availability (arrival time, urgency) and data properties. As a sequence the unification of ways of representation and use the data in the data scenarios with incompleteness and uncertainties is required. The use of different scenarios of data processing in DSS leads to the result of different accuracy and completeness which can be obtained in different time. Various scenarios are not mutually exclusive, or blocking a joint performance and they can be initiated by the users according to the development of the analyzed situations and decision-making process. This is connected with the inevitable problems of competition for computing resources, which raises the problem of the performance providing for data processing in the DSS. To solve it apply technology of distributed systems, including cloud computing, are required.
DSS based on operational processing of data from external sources can be effectively implemented subject to a number of principles and solutions delivered by universal platform CLAVIRE. These include the implementation of a single, integrated data source with the unification of formats, statistical control, restoring of missing values and correction of errors, and a flexible choice of workflows depending on the completeness and the characteristics of the incoming data. Platform CLAVIRE provides the necessary infrastructure for distributed launch of services and implementation of computations on a set of measured data from the user and historical data. This approach was successfully used in development of early warning system for flood prevention in Saint-Petersburg though the general approach for development of such kind of systems allows replacing the computational services and adapting them to other application areas.