FMCache: File-System Metadata Caching in Programmable Switches

2025-10-09

AI Summary

To address poor metadata access scalability, high cache coherence overhead, and server overload in distributed file systems under multi-client workloads, this paper proposes FMCache—the first programmable switch-based metadata caching scheme with path awareness. FMCache innovatively leverages the Tofino switch’s data plane for fine-grained file path parsing and metadata caching, overcoming the semantic blindness of conventional key-value caches to hierarchical paths. It further introduces a lightweight coherence protocol that guarantees strong consistency under stringent switch resource constraints. Evaluated on HDFS using real-world workloads, FMCache achieves up to 181.6% higher throughput than native HDFS; even when layered atop existing client-side caching, it delivers an additional 139.6% throughput gain. Moreover, it incurs low latency and maintains bounded, predictable resource consumption on the switch.

Technology Category

• Search and Optimization: Distributed Search
• Data Mining & Knowledge Management: Scalability, Parallel & Distributed Systems
• Machine Learning: Hardware-aware ML

Application Category

• Security and Privacy: Large-scale security measurements
• Systems and Infrastructure for Web, Mobile and WoT: Data management and stream processing for Web, mobile and wireless applications
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems

Abstract

Fast and scalable metadata management across multiple metadata servers is crucial for distributed file systems to handle numerous files and directories. Client-side caching of frequently accessed metadata can mitigate server loads, but incurs significant overhead and complexity in maintaining cache consistency when the number of clients increases. We propose FMCache, an in-switch file-system metadata caching framework that leverages programmable switches to serve file-system metadata requests from multiple clients directly in the switch data plane. Unlike prior in-switch key-value caching approaches, FMCache addresses file-system-specific path dependencies under stringent switch resource constraints. We implement FMCache atop Hadoop HDFS and evaluate it on a Tofino-switch testbed using real-world file-system metadata workloads. FMCache achieves up to 181.6% higher throughput than vanilla HDFS and complements client-side caching with additional throughput gains of up to 139.6%. It also incurs low latencies and limited switch resource usage.

Problem

• Distributed file systems struggle with scalable metadata management
• Client-side metadata caching causes high consistency overhead
• Programmable switches enable efficient in-network metadata caching

Innovation

• Programmable switches directly serve metadata requests
• Addresses file-system path dependencies in switches
• Achieves high throughput with low switch resource usage