π€ AI Summary
Distributed FUSE file systems have long struggled to simultaneously achieve write-back page caching and strong consistency: existing approaches either sacrifice consistency for performance or rely on inefficient cache mechanisms. This paper introduces DistFUSEβthe first FUSE-based distributed file system supporting kernel-level write-back page caching while guaranteeing strong consistency. Its core innovation is offloading distributed lock management to the kernel driver, enabling coordinated cross-node page cache access and eliminating local blind updates. DistFUSE is the first to integrate write-back caching with strong consistency semantics in FUSE, underpinned by a POSIX-compliant, multi-node cooperative caching protocol. Experimental results demonstrate that, under strict strong consistency, DistFUSE achieves up to a 75% throughput improvement over state-of-the-art baselines, significantly breaking the performance bottleneck inherent in existing FUSE-based distributed file systems.
π Abstract
The large-scale, multi-tenant nature of cloud computing requires distributed file systems that offer stability, adaptability, and compatibility. FUSE-based distributed file systems have emerged as a popular solution for the cloud, offering fast deployment, fault isolation, and POSIX compliance. However, FUSE's performance limitations, particularly its inability to reconcile page caching with strong consistency in distributed environments, remain a persistent problem. Existing approaches either sacrifice consistency for performance or rely on inefficient caching, limiting their practicality. To this end, we present DistFUSE, the first FUSE-based distributed file system that relies on a write-back kernel-based page cache for performance and provides strong consistency. DistFUSE achieves this by offloading userspace lock management to the kernel driver, allowing coordinated access to the kernel's page cache across nodes. This design eliminates blind local cache updates and ensures cluster-wide consistency without compromising performance. Our evaluation shows DistFUSE improves throughput by up to 75% compared to baseline approaches.