ScaleDefrag: Design and implementation of a scalable file defragmentation tool for flash-based SSDs

Area of Science:

• Computer Science
• Data Storage Systems

Background:

• Flash-based Solid-State Drives (SSDs) are prevalent in cloud and High-Performance Computing (HPC).
• File fragmentation on SSDs, caused by intensive file modifications, degrades storage system performance.
• Existing defragmentation tools lack scalability due to sequential and synchronous processing.

Purpose of the Study:

• To develop a scalable defragmentation tool for flash-based SSDs.
• To reduce defragmentation time and improve system performance.

Main Methods:

• ScaleDefrag employs parallel and asynchronous processing for defragmentation.
• An information collector gathers file metadata prior to parallel execution.
• Multiple defraggers work on a one-to-one file basis with asynchronous I/O operations.

Main Results:

• ScaleDefrag reduces defragmentation time by up to 3.8× compared to e4defrag.
• The tool demonstrates improved scalability on multi-core systems.
• ScaleDefrag increases defragmentation throughput and mitigates performance degradation for co-running applications.

Conclusions:

• ScaleDefrag offers a scalable and efficient solution for SSD defragmentation.
• The parallel and asynchronous approach overcomes the limitations of traditional defragmentation methods.
• This enhances the performance and reliability of large-scale storage systems utilizing SSDs.