OPERATION OF UNMANAGEABLE INFORMATION DATA CENTER AT HIGH NUMBER OF LINKS WITH DATA TERMINALS

Abstract:

An important step in development of science infrastructure is support of operation for unmanageable (assumed no direct human involvement) data centers (DC). These DC can automatically handle a big volume of data through the high (unlimited) number of links with independent data terminals (DT). Present initiative considers theoretical model for evolution of DC at assumption of unlimited access to the bidirectional data flow y between DC and DT. Major element of this model is execution of the data link DT ↔ DC through a formalism of continuity equation.

Result of above research is evolutionary curve demonstrating dependence of average efficiency of data exchange ϒ on y (Fig.1). This dependence shows the non-linear nature of evolution process. Special interest of this dependence is presence of two dissimilar operational phases. While in phase (1) operation of DC takes the form of data accumulation, in the phase 2 we observe appearance of the data ordering with arising of data clusters grouped around the most stable data links at the expense of riding of the redundant data which manifests in the discreteness of spectrum for ϒ and y. Besides, spectrum of the informational entropy S also becomes discrete in the limited y-range. Further in evolution (subject of available data flow y), in the phase 2 operation of DC can go through the counted number of critical points in which the reconfiguration of information exchange is expected. In addition, in phase 2 operation of DC demonstrates unique features which do not exist in the phase 1. We mean prevalence of the individual behavior over the collective one, organizing over chaotic, appearance of the dynamic data archive, and exchange with it through the stable data links. So, effective operation of unmanageable DC with the high number of DT can go through the forming of predefined phases. Awareness of these phases may help to proper monitoring for status of DC operation.

Capture.

Fig.1. Evolution of operation for data center with unlimited number of links with data terminals. In the plot, by an abscissa axis a data exchange flow y is indicated, by the ordinate axis an average efficiency of data exchange ϒ is indicated. Two major phases are shown – the data accumulation (marked in dark grey) and the data ordering (marked in light grey). Curve shows growth, maximization (indicated by dash) and decaying of ϒ.