While Miss Ratio Curve (MRC) profiling methods based on spatial sampling are effective in modeling cache behaviors, previous MRC studies lack in-depth analysis of profiling errors and primarily target homogeneous object-size scenarios. This has caused imbalanced errors of existing MRC approaches when employed in heterogeneous object-size caches. For instance, in CDN traces, the error of the Byte Miss Ratio Curve (BMRC) could be two orders of magnitude larger than that of the Object Miss Ratio Curve (OMRC).

In modern cloud block storage systems, the routing layer bears the task of slicing and distributing application requests to the underlying storage engine and thus becomes the critical path. However, skewed application traffic leads to load imbalance among routing servers, causing severe performance bottlenecks and hurting the quality of service. Existing load balancing methods typically employ static allocation strategies without considering the access pattern of individual virtual disks. They are not well suited to deal with the highly dynamic loads in production. In this paper, we first collect and analyze 7-day workload traces of 130k virtual disks from a commercial cloud block storage system. Based on several observations, we propose VDMig, an adaptive virtual disk migration scheme. VDMig initially divides virtual disks into two types of access patterns, then uses application-level semantics to characterize the load of virtual disks, and finally implements predictive migrating strategies adaptively to achieve dynamic load balancing. Extensive experiments demonstrate that VDMig can achieve fine-grained virtual disk management, effectively balancing the load among routing servers. Compared to the static methods widely deployed in the industry, VDMig can reduce the imbalance by 82.38 % on average.