System and method for dynamically power and performance optimized server interconnects

Abstract

A system and method for packet switching functionality focused on network aggregation that reduces size and power requirements of typical systems are provided in which the system and method also increases bandwidth and reduces latency from typical deployed systems.

What is claimed is: 
     
         1 . A method comprising:
 configuring, via a fabric management unit of a node, a fabric switch, wherein the fabric switch includes a plurality of links to a plurality of ports;   receiving interrupts at the fabric management unit; and   determining, at the fabric management unit, an adjustment of link rates for the plurality of links in the fabric switch based, at least in part, on the received interrupts.   
     
     
         2 . The method of  claim 1 , wherein the interrupts are received based on reception and transmission of control messages from the fabric management unit. 
     
     
         3 . The method of  claim 2 , wherein the control messages bypass a transmit buffer of the node. 
     
     
         4 . The method of  claim 2 , wherein each link further comprises a plurality of lanes over which data is communicated, and wherein the control messages are distributed over each lane. 
     
     
         5 . The method of  claim 1 , further comprising measuring a statistic regarding each link, wherein the adjustment of link rates is based, at least in part, on the measured statistic. 
     
     
         6 . The method of  claim 5 , wherein the statistic is measured at a receive channel and at a transmit channel. 
     
     
         7 . The method of  claim 5 , wherein the statistic is an arbitration score indicative of how long a receive channel has been waiting to win arbitration of the fabric switch. 
     
     
         8 . The method of  claim 5 , wherein the statistic is a bandwidth utilization of each link. 
     
     
         9 . The method of  claim 1 , wherein the fabric switch contains a set of rules to control the link rates of the plurality of links. 
     
     
         10 . The method of  claim 9 , wherein the rules comprise one of a configuration of a bandwidth, a configuration of the speed of a link, a control of the speed of a transmit channel and a receive channel of the link, a maximum power used by the switch, or a dynamic routing of traffic. 
     
     
         11 . The method of  claim 1 , wherein each link comprises a receive channel having one or more receive lanes and a transmit channel having one or more transmit lanes, wherein the method further comprises:
 determining a number of active receive lanes that receive data;   determining a number of active transmit lanes that transmit data;   determining a receive rate of the active receive lanes; and   determining a transmit rate of the active transmit lanes.   
     
     
         12 . The method of  claim 11 , wherein the number of active receive lanes is different than the number of active transmit lanes. 
     
     
         13 . The method of  claim 11 , further comprising:
 receiving a message on a receive channel;   determining the received message is a control message;   removing the control message from the receive channel; and   interrupting a processor based on receiving the control message.   
     
     
         14 . A non-transitory computer-readable medium having instructions stored thereon, the instructions comprising:
 instructions to configure a fabric switch at a fabric management unit of a node, wherein the fabric switch includes a plurality of links to a plurality of ports;   instructions to receive interrupts at the fabric management unit; and   instructions to determine, at the fabric management unit, an adjustment of link rates for the plurality of links in the fabric switch based, at least in part, on the received interrupts.   
     
     
         15 . The non-transitory computer-readable medium of  claim 14 , wherein the interrupts are received based on reception and transmission of control messages from the fabric management unit. 
     
     
         16 . The non-transitory computer-readable medium of  claim 15 , wherein the control messages bypass a transmit buffer of the node. 
     
     
         17 . The non-transitory computer-readable medium of  claim 15 , wherein each link further comprises a plurality of lanes over which data is communicated, and wherein the control messages are distributed over each lane. 
     
     
         18 . The non-transitory computer-readable medium of  claim 14 , further comprising instructions to measure a statistic regarding each link, wherein the adjustment of link rates is based, at least in part, on the measured statistic. 
     
     
         19 . The non-transitory computer-readable medium of  claim 18 , wherein the statistic is measured at a receive channel and at a transmit channel. 
     
     
         20 . The non-transitory computer-readable medium of  claim 18 , wherein the statistic is an arbitration score indicative of how long a receive channel has been waiting to win arbitration of the fabric switch.