MES-713: Parallelize Fuse Interface

[markovejnovic]

No description provided.

[mesa-dot-dev]

Mesa Description

TL;DR

This PR parallelizes the FUSE request processing loop, allowing multiple filesystem operations to be handled concurrently. This significantly improves filesystem throughput and responsiveness, especially under heavy I/O load.

Why we made these changes

The previous FUSE implementation was single-threaded, processing all filesystem requests serially. This created a performance bottleneck, causing slow and unresponsive behavior when multiple applications or processes accessed the filesystem simultaneously (e.g., during builds, searches, or Git operations). By parallelizing the request handling, we can remove this bottleneck and dramatically improve performance.

What changed?

• The main FUSE server loop was refactored to dispatch incoming requests to a worker thread pool instead of handling them sequentially.
• Filesystem data structures, including the inode cache (icache) and cloud-backed file system components (mescloud), were made thread-safe using appropriate concurrency primitives to prevent race conditions.
• The core filesystem trait was updated to ensure all implementations can safely handle concurrent requests.

Validation

• All existing integration and unit tests pass, ensuring no regressions in filesystem correctness.
• Ran stress tests with high levels of concurrent I/O to verify stability and check for deadlocks or race conditions.
• Benchmarked key filesystem operations (e.g., directory listings, file reads) to confirm significant performance gains under load.

^{Description generated by Mesa. Update settings}