In this paper, an analytic reliability algorithm, in which the reliability indices such as ENS and PNS are expressed as polynomials of 0 1 switch variables, is proposed firstly.

This partitioning of data into different clusters allows them to ensure high reliability for critical jobs, while still allowing the use of Hadoop for more exploratory and analytic purposes.

The errors from outside interference are reduced and reliability of the analytic result is enforced by performing analysis, process and determination of the internal relations among the three signals.