Method of improving over protocol-required scheduling tables while maintaining same

摘要

Public scheduling tables of PCI-Express network devices are remapped into private scheduling tables having different data structures. The private scheduling tables enable the construction of parallel-processing selection engines (ones with look-ahead selection capabilities) that are more compact in size than would have been possible with use of the data structures of the public scheduling tables. In one embodiment, residual weight values are re-shuffled so as to move each winner of an arbitration round away from a winner's proximity bar by a distance corresponding to an initial weight assigned to the winner. The initial weight can be proportional to the reciprocal of a bandwidth allocation assigned to each data source.