| 摘要 |
A data network includes a switch that links a number of host processors to shared resources such as network ports of a cached storage subsystem. The switch routes resource requests from its inputs to selected ones of its outputs in accordance with programmed routing information. The routing information is changed dynamically based on measured loading characteristics of requests from the switch inputs or the respective hosts, in order to reduce the probability of blocked paths. In a specific example, respective subsets of output ports of the switch are assigned to each respective input port, the loading of each input port of the switch is measured, and the respective subset of output ports assigned to each input port are repetitively changed, and the respective subset of an input port is changed in response to blocking of the output ports assigned to the input port. A respective subset is selected based on the measured loading characteristics and a priority assigned to the input port, in order to balance loading on the output ports. In another specific example, a respective subset of the output ports is assigned to each host, based on host loading characteristics and a respective priority assigned to each host.
|
