Quote:
Originally Posted by Chrysalis
interesting info on the timeslots, so it is possible I guess to get jitter even when the bandwidth itself isnt saturated but if too many people want a timeslot at the same time.
I too learn from seph.  |
Unsure what you mean there - could you elaborate? If there are too many outstanding requests for mini-slots that is the definition of saturation on an upstream. Utilisation on upstreams isn't measured in bandwidth terms it's actually measured in % of mini-slots used.
Modem gets data from customer's equipment, requests the appropriate amount of mini-slots. There is the potential for minimal jitter, on the order of a millisecond or two, where the modem sends a request and it collides with a request from another modem on the way up to the CMTS but this is not a significant source of jitter. The major cause is where the CMTS has no mini-slots available for a while, or indeed at all, in this case it'll either give the modem a zero-length grant and the modem will have to re-request the bandwidth again or it'll just not give any response at all.
Where a modem has a 100ms delay it has gone through repeated cycles of requesting bandwidth and receiving zero length grants or none at all until eventually, nearly 100ms later, it finally receives a grant and transmits its' data.
Where a modem shows packet loss as well a buffer in the modem filled up and overflowed, so the data in there for the longest period was dropped to make room for newer data. If congestion is heavy enough it's quite possible for this to happen as the modem will only hold so much data while waiting for upstream grant.