Lustre File System SSD Caching for Peak Performance

The patent segments the storage system into two distinct parts: a small, fast cache layer using solid-state drives (SSDs) for frequently accessed data, and a large, slower storage layer using rotating disks for bulk data. This segmentation allows the system to achieve high peak performance through the SSD cache while avoiding the need to provision a large number of rotating disks for peak loads, thus resolving the contradiction between performance and device complexity.

The patent changes the performance parameter of the storage system by introducing SSDs with significantly higher I/O speeds compared to traditional rotating disks. By using SSDs for the cache layer, the system achieves peak performance requirements without needing to scale out the number of rotating disks, thereby reducing device complexity and resource utilization costs while maintaining high productivity.

The storage system is segmented into an SSD cache layer and a rotating disk storage layer. The SSD cache handles peak performance demands with low power consumption, while the rotating disks handle bulk storage with higher capacity but are not needed at full capacity for peak performance. This segmentation reduces overall power consumption compared to using only rotating disks to meet peak performance requirements.

The patent uses SSDs as a temporary cache layer that holds frequently accessed data. These SSDs are used intensively during peak performance periods and then data is flushed back to the rotating disks. The SSD cache acts as a short-term holding area that enables peak performance without requiring permanent provisioning of a large number of high-performance rotating disks, thereby reducing power consumption and space requirements.

The patent segments the storage hierarchy into an SSD cache layer with smaller capacity but higher performance, and a rotating disk storage layer with larger capacity but lower performance. The SSD cache layer is sized to hold the working set of frequently accessed data, providing high caching performance. The rotating disk layer provides the bulk storage capacity. This segmentation resolves the contradiction by allowing the system to achieve high caching performance with a limited SSD capacity while maintaining large overall storage capacity through the rotating disks.

The patent changes the capacity parameter by using SSDs for the cache layer where high I/O performance is critical rather than maximizing capacity. The SSD cache is optimized for speed with smaller capacity, while the rotating disks provide the large-capacity bulk storage. This parameter change allows the system to achieve high caching performance without sacrificing overall storage capacity, as the SSD cache works in conjunction with the larger-capacity rotating disk storage.