In order to change the status that distributed storage system can’t adapt to dynamic workloads, we introduce a new concept of evolution redundancy in logistic evolution storage system. It associates the redundancy level that characterizes the storage organization to sets of data and not to storage areas while eliminating data migration for reconfiguration. Inside storage node, the storage space is divided into some sequential region that can automatically adjust data layout on the basis of workload without data migration. Then evolution redundancy makes the map between logic block address and physical block address become very simply so as to avoid bottleneck incurred by complex map. For saving storage space during transforming between various RAID levels, we propose independent disks group by which our system not only implement evolution redundancy but also achieve well performance/cost. Finally, we carry out a large number of simulations to validate transformation threshold between different RAID levels.
展开▼
