NetCAS: Dynamic Cache and Backend Device Management in Networked Environments

TL;DR

NetCAS tackles the problem of efficiently using both cache and remote backend storage in disaggregated datacenter environments where network-induced performance variability can undermine static caching strategies. It introduces a Perf Profile–based dynamic split model coupled with real-time network feedback and an inline Batched Weighted Round Robin scheduler to enforce the split with minimal overhead. The approach yields substantial improvements, including up to 174% higher throughput versus traditional caching and up to 3.5× gains over converging baselines like OrthusCAS under fluctuating network conditions, while maintaining transparency and low CPU overhead. This work provides a scalable blueprint for network-aware, hybrid storage systems in modern datacenters, with potential extensions to richer monitoring and mixed read/write workloads.

Abstract

Modern storage systems often combine fast cache with slower backend devices to accelerate I/O. As performance gaps narrow, concurrently accessing both devices, rather than relying solely on cache hits, can improve throughput. However, in data centers, remote backend storage accessed over networks suffers from unpredictable contention, complicating this split. We present NetCAS, a framework that dynamically splits I/O between cache and backend devices based on real-time network feedback and a precomputed Perf Profile. Unlike traditional hit-rate-based policies, NetCAS adapts split ratios to workload configuration and networking performance. NetCAS employs a low-overhead batched round-robin scheduler to enforce splits, avoiding per-request costs. It achieves up to 174% higher performance than traditional caching in remote storage environments and outperforms converging schemes like Orthus by up to 3.5X under fluctuating network conditions.

Figures (8)

• Figure 1: Throughput comparison between cache device (PMem), backend device (NVMe), and splitting at the optimal ratio across varying thread counts. The percentage labels on the splitting line denote the optimal split ratio at each concurrency (e.g., 75% indicates 75% of requests sent to cache and 25% to backend).
• Figure 2: NetCAS framework overview. Real time metrics from NetCAS Monitor and device baseline performance from NetCAS Perf Profile are passed to NetCAS Splitter, where I/O requests are dynamically routed to the local cache and remote backend device.
• Figure 3: Break-even (BE) analysis for NetCAS (inflight requests = 16, threads = 16). The full table was constructed from a grid of 5 inflight levels $\times$ 5 thread levels $\times$ 2 block sizes with 30 s per point, requiring about 25 minutes for the one-time build.
• Figure 4: Normalized throughput under different inflight request counts without network congestion. At low concurrency the calculated split deviates from the empirical best, but accuracy improves quickly with higher concurrency, converging to the optimal ratio.
• Figure 5: Throughput comparison of BWRR versus random dispatch across inflight requests and thread counts. BWRR sustains the target ratio more evenly, delivering higher aggregate IOPS especially under shallow queues where randomization causes imbalance.