Techniques implemented in KPC ToolBox

KPC ToolBox is built upon the characterization results of a general class of Markovian Arrival Processes(MAP). Intuitively, a MAP works like an Finite State Machine(FSM) with n states. State transitions happen from one state to another, some of which may trigger the events of arrival while some others don't. Define a random variable X as the time between two consecutive arrival events. For a real trace, the random variable X defines the inter-arrival times in the traffic of systems like networking, disks and etc.

KPC ToolBox fits the trace of inter-arrival times and generate a MAP. There are two novel fitting techniques implemented in this ToolBox. First, a recently-proposed Kronecker Product Composition(KPC) fitting method. The basic idea is to reduce the moment and temporal dependence fitting problem to assigning the characteristics of smaller MAPs composed by no more than 2 states. The second innovation of the KPC-Toolbox is the automatic determination of the order of the MAPs used in fitting(the number of the states in the "FSM"), while order selection remains as an open problem in the literature of MAP fitting.

KPC ToolBox incorporates these two innovative techniques and now is open to Public for testing. It can be found in the Download page.