re>>:"...and Cable Modem approaches, which are entirely contention based. And where some cable modem approaches are slightly different than pure contention based, they nonetheless vye for the same downstream bandwidth even though there may be arbitration tricks taking place somewhere in the head end."
First, I assume you meant "upstream" since there is no contention in the downstream. It is scheduled, but there is no access contention. There is only one accessing station (the CMTS) so contention is not an issue. And in general, since proprietary cable modems are both fading away, and not publicly documented as to their operation, we can skip them in this discussion (but there are almost certainly proprietary modems that are mostly contention based). DOCSIS, on the other hand, is NOT mostly contention based. It falls more into the PDAMA (packet demand assigned multiple access) camp, where contention transmits are only used for requesting a reservation (and are very small packets--something like 6 bytes). The bulk of the upstream traffic, percentage wise, is in granted (non-contention)timeslots. Going forward with DOCSIS 1.1, contention slots should--in theory--become an even smaller percentage of upstream bandwidth utilization since there is a lot of flexibility in pre-assigned or dynamically assigned traffic shapes which utilize unicast polling or pre-arranged periodic grants that don't require any reservation bandwidth whatsoever.
But hey, efficiency can only take you so far. Much greater bandwidth is required, ultimately. |
