Pipette: Efficient Fine-Grained Reads for SSDs
TimeTuesday, July 12th3:50pm - 4:10pm PDT
Location3004, Level 3
Event Type
Research Manuscript
Embedded Memory, Storage and Networking
Embedded Systems
DescriptionWe present Pipette, an efficient fine-grained reads framework, to effectively exploit byte-addressable SSDs to address the challenge of varying request sizes. We extend the virtual memory management to adaptively handle read requests through both block I/O interface and byte-addressable interface. Besides, we design an adaptive fine-grained cache to fully utilize memory resources and gain benefits from the fast access to host-side DRAM, while preventing memory from fragmentation. Evaluation results show that Pipette outperforms the state-of-the-art in improving read performance and reducing I/O traffic.