MCU Pool Orders MCU Management
Polycom, Inc. 161
Every 30 minutes, the reliability score of the MCU is increased so that it won’t
be permanently removed from the pool due to failures in the distant past. To
avoid trying the MCU every 30 minutes, monitor the DMA system and
administratively take the MCU out of service.
By increasing the number of MCUs in the pool or increasing their capacity, you
can decrease the usage of the working MCUs during a failover scenario. So, for
example, if you want to avoid routing any more calls to an MCU after two
consecutive failed calls, provide enough excess capacity that the remaining
MCUs never all reach 43% port usage during a failure.
See also:
“MCU Pool Orders” on page156
“MCU Selection Process” on page159
“MCU Pool Order Procedures” on page162
784%
888%
990%
Table 6-15 How failed calls reduce the apparent capacity of an MCU (continued)
Consecutive
Failed Calls Percentage of Ports
Assumed To Be in Use
Note
After each call, the DMA system recalculates the reliability of an MCU as the
weighted average of the result for the current call (1 for success, 0 for failure) and
the reliability of all previous calls, using this formula:
reliability = (current_call + (weight * previous_reliability)) /
(1 + weight)
For example, if weight is 5, previous reliability is 1 (no previous failed calls), and the
call is sucessful, the reliability remains 1:
(1 + (5 *1)) / (1 + 5) = 1
If weight is 5, previous reliability is 1, and the call fails, the reliability becomes 5/6:
(0 + (5 *1)) / (1 + 5) = 5/6
If weight is 5, previous reliability is 5/6, and the call is successful, the reliability
becomes 31/36:
(1 + (5 *5/6)) / (1 + 5) = 31/36
If the reliability is ever less than 1, it exponentially approaches 1 as more calls
succeed, but it never quite gets there. It very quickly reaches the point where the
weight of the past failed call counts less than a single call in progress. But it remains
as the tie breaker between completely unused MCUs forever.