Load Switching
remote is divided by the total Bytes in Queue for all remotes to give the percentage allocation for each remote.)
•Dynamic Slot Mode - The slot size for each remote is computed based on the time (at the current data rate) needed to transmit all the Bytes in Queue. If the result is less than the minimum slot size or more than the maximum slot size, the slot is adjusted accordingly.
•GIR (Guaranteed Information Rate) Mode - Initially computed the same as Dynamic Cycle except there is no maximum limit. After all remotes have been assigned slots, the burst map is checked to see if the total cycle length exceeds 1 second. If not, then all requirements are satisfied and the burst map is complete. However, if the cycle is greater than one second, then the slots are adjusted proportionally so that all remotes receive at least their guaranteed rate plus whatever excess is still available. (In the current design, when the 1 second restriction is exceeded, remotes without a specified GIR are reduced to the global minimum slot size and the remaining bandwidth is distributed amongst remotes that have been assigned a GIR rate. This approach is based on the assumption that remotes that have been assigned a GIR are paying a premium and should benefit from available excess bandwidth when needed. Note that the GIR allocations are restricted so that the assigned GIR totals cannot exceed available bandwidth. If this restriction is somehow violated, then it will not be possible to properly allocate bandwidth when the network is overloaded.)
•Entry Channel Mode - This is the same as Dynamic Cycle, except that as soon as the Hub receives an STDMA ACK, it initiates a switch to SCPC mode based on the policy set for that remote.
The important thing to understand about “Bytes in Queue” is that any data that is not transmitted (i.e. does not fit) in the next slot will be reported again in the next STDMA ACK. Thus the “Bytes in Queue” is not necessarily an accurate measure of the actual traffic being passed through the remote.
The “Queued Bytes” on the other hand, reflects only the data that was received in the last cycle and thus is never duplicated (not including TCP retransmis- sions). This is the metric that is used for computing average load and initiating a load switch as needed.
Before discussing how load switching is determined, it is necessary to explain the user parameters that control the switch. The menu shown in figure
Vipersat |
