Why Multi-Agent is so important in risk management application

All the calculations for risk management application are very complicated and resourrce hungry. We need to calculate risk of company portfolio, then the portfolio of each department, then the portfolio of each desk, then the portfolio of each trader. Fortunately these calculations can be separated and calculcated independently. That's why a multi-agent system or concurrent environment is a must to build a risk management system.

Scope of namespace RiskLib.MultiAgent in Release 1.0

A General-Purpose Multi-Agent System

To define a general-purpose multi-agent system can be used in any kinds of calculation or job.

Registration and Activity Monitor of Agent

Registration, messaging, and logging is provided by providers which are defined by users of this library. This library defines only objects and behaviors of providers. User can define providers by using databases, web services, or file systems. Agent should report in a given frequency to keep itself alive.

Centralized Task Allocation

Task is labeled by a numeric field Priority. All agents request task from a centralized coordinator, defined in providers, and report back the status. The coordinator response with tasks with the highest priority and mark it as assigned. Assigned task can be reassigned after timeout or agent died.

Extensible Task Data Type

The generic task data type has no knowledge about the data structure of parameters or procedures to be performed. User can extend the Task object to do jobs not defined in this library.

Last edited Apr 27, 2009 at 8:28 PM by alvincho, version 2


No comments yet.