Abstract:
The load balancing has always been a research hotspot in the field of storage systems, but there are still shortcomings in its application and implementation in distributed object-based storage systems.Aiming at the facts that load balancing in object-based storage system mostly relies on experience or limited dynamic configuration to guide the execution of the algorithm, the migration rhythm cannot be controlled during the load rebalancing process, the algorithm implementation is complex, and the real-time load conditions of storage nodes are not fully considered, a dynamic load balancing method based on monitoring is proposed. In this method, the monitoring system is introduced into the object-based storage system to complete the real-time perception of the load status of each storage node; the migration algorithm is triggered by comprehensive estimation according to the node load, heterogeneity and other factors, avoiding relying on experience and preset values to guide the execution of the algorithm; Use the flexible custom event trigger mechanism and fault automatic processing mechanism of the monitoring system to design a load migration strategy to dynamically complete the load rebalancing between nodes, and reduce the complexity of the algorithm implementation; Use the multi-step operation sequence flow control in the monitoring system action execution process to complete one-by-one migration, and effectively control the load migration rhythm. The results of the verification of the method applied to the self-developed object-based storage system CSCloud show that the method can obtain the load information of each node in the cluster in a timely manner and achieve timely and accurate dynamic load balancing, which provides a new idea for the dynamic load balancing of object-based storage system and other distributed storage systems.