Teleport: Load Balancing with Dynamic Adjustment Strategy Based on OSD Storage System

The development of object-based storage devices (OSDs), widely regarded as a promising storage standard, is sprouting up these days. In this paper, we bring forward a new framework of resource allocation to achieve load balancing for OSD storage systems. Compared with traditional researches that schedule requests passively, we exploit the intelligence of object storage target (OST) to generate local resource information, which is combines with the global information synthesized by the Metadata Server (MDS) to balance workloads and tune the performance. To attain this goal, a dynamic adjustment strategy is proposed to calculate some resource-related factors by MDS as well as OST, which guide each OST to adjust the data distribution and assign which OST should process a new request. Moreover, how to extend the object attributes of OSD standard to achieve this strategy is also introduced. At last, the experimental results demonstrate that this method can balance the load among OSTs fairly.