6. Algorithm Rating

The series of experiments conducted before have already verified the excellent performance of AvgShedder:

• In heterogeneous environments, AvgShedder attempts to assign the heaviest load to the broker with the highest performance. This contrasts with UniformLoadShedder + LeastLongTermMessageRate, which unloads the bundles from high - performance brokers onto low - performance ones.

• Even when confronted with complex scenarios such as load fluctuations, the system can maintain stable operation. In our production environment, only one bundle unload operation was triggered within a month, whereas ThresholdShedder executed bundle unload operations tens of times in a single day.

• AvgShedder does not have the over placement and over unloading problems faced by ThresholdShedder + LeastResourceUsageWithWeight, nor does it issue incorrect load balancing commands.

• The load - balancing speed is remarkably rapid. In the majority of cases, merely one or two bundle unload operations are required for the cluster to reach a stable state.

Consequently, we have the following scoring table:

Strategy	Adaptability to Heterogeneous Environments (Adaptability)	Adaptability to Load Fluctuations (Stability)	Over Placement (Correctness)	Over Unloading (Correctness)	Speed
ThresholdShedder + LeastResourceUsageWithWeight	Fair	Good	Poor	Poor	Fair
UniformLoadShedder + LeastLongTermMessageRate	Poor	Poor	Good	Good	Fair
AvgShedder	Fair	Good	Good	Good	Good