Abstract: A system and method for Quality of Service (QoS) based server cluster power management is disclosed. The method of the present invention includes: grouping activities within a server cluster into predefined sets; assigning a priority level to each set; identifying a server hosting a set of lower-priority activities within the cluster; receiving a power interruption signal; and diverting power reserves of the server to another server in the cluster, in response to the power interruption signal. The system of the present invention includes: servers, hosting a plurality of activity sets each having an associated QoS level; power reserves coupled to the servers; a switch matrix coupled to direct the power reserves between the servers; and a power manager, coupled to the switch matrix, for commanding the switch matrix to divert power from servers hosting low QoS activity sets to servers hosting high-priority activity sets, in response to a power interruption.

Abstract: A system and method for Quality of Service (QoS) based server cluster power management is disclosed. The method of the present invention includes: grouping activities within a server cluster into predefined sets; assigning a priority level to each set; identifying a server hosting a set of lower-priority activities within the cluster; receiving a power interruption signal; and diverting power reserves of the server to another server in the cluster, in response to the power interruption signal. The system of the present invention includes: servers, hosting a plurality of activity sets each having an associated QoS level; power reserves coupled to the servers; a switch matrix coupled to direct the power reserves between the servers; and a power manager, coupled to the switch matrix, for commanding the switch matrix to divert power from servers hosting low QoS activity sets to servers hosting high-priority activity sets, in response to a power interruption.

Abstract: A method and system of determining circuit performance-related characteristics, particularly delay and crosstalk, of interconnects includes defining a number of process variables which exhibit Gaussian distributions with respect to geometrical variances. A table of statistically based worst-case values representative of capacitances and resistances associated with different selections of the process variables is generated. In the preferred embodiment, the statistically based worst-case values are 3-sigma values. Also generated is a table of capacitance derivatives with respect to interconnect geometries. When a particular interconnect layout having selected process parameters is designated, the tables of 3-sigma values and derivatives are accessed to generate a resistance-capacitance (RC) net representative of the interconnect layout. The resistance and capacitance are correlated for each RC net and are partially determined by a randomized selection of values for geometries of the interconnect layout.