In-Memory Cache Extent Pre-Allocation for Snapshot Latency

The patent pre-allocates extents in an in-memory cache before snapshot operations are needed. By identifying and reserving the required extents in advance within the cache structure, the system eliminates the need for expensive on-disk allocation operations during actual snapshot creation, thereby reducing system load while maintaining continuous snapshot capability for RPO reduction.

The patent introduces an in-memory cache as an intermediary between the snapshot operation and the physical storage allocation. This cache layer absorbs the allocation/freeing operations, providing a buffer that reduces the impact on the main storage system. The cache acts as a mediator that handles extent management internally, preventing direct I/O operations during snapshot creation and deletion.

The patent segments the extent management into two parts: an in-memory cache for active allocation operations and an on-disk persistence layer for durability. The in-memory cache handles frequent allocation/freeing operations locally without serialization, while only committing changes to disk when necessary. This segmentation separates the high-performance operational layer from the persistence layer, eliminating the need for serialization during normal operations.

The patent performs extent identification and allocation decisions in-memory before any disk I/O occurs. By pre-computing which extents need to be allocated and managing them in the cache structure, the system avoids repeated serialization and disk access operations. The preliminary in-memory processing reduces the frequency and impact of disk serialization operations.

The patent implements a self-managing in-memory cache that automatically tracks and manages extent allocations and freeings without requiring extensive CPU intervention. The cache structure maintains its own state information about allocated and available extents, enabling it to service snapshot operations autonomously. This self-service capability reduces CPU load while maintaining the ability to create and delete snapshots frequently for freshness.

The patent changes the state management parameter from disk-based to memory-based. By moving extent tracking from the disk to an in-memory cache, the system fundamentally alters how allocation and freeing operations are handled. This parameter change enables frequent snapshot operations to be processed in-memory with minimal CPU overhead, as the cache can quickly identify and allocate extents without expensive disk access or complex serialization.