Multi SSD Management Segregating Garbage Collection from Writes

The patent divides the SSD into multiple independent blocks, where each block is dedicated to storing pages of a specific virtual machine. This segmentation prevents the need to manage pages of different VMs within the same block, eliminating the time loss associated with managing mixed VM pages during garbage collection. When garbage collection is needed for a specific VM, only its dedicated block needs to be processed, not blocks containing pages from other VMs.

The patent implements a wear-leveling mechanism that performs preliminary actions to distribute write operations evenly across different SSD blocks before any block becomes fully worn out. By proactively managing block usage and performing wear-leveling migrations before performance degradation occurs, the system maintains consistent write performance over time. The controller tracks block usage and proactively redistributes data to prevent any single block from being overused, thus maintaining performance consistency.

The patent segments SSD resources into dedicated blocks for each virtual machine, allowing resource sharing at the system level while maintaining isolation at the block level. Each VM has its own dedicated blocks that are not accessed by other VMs, preventing performance interference. The controller manages the allocation and migration of these dedicated blocks, enabling flexible resource sharing while ensuring that garbage collection and write operations for one VM do not impact the performance of other VMs.