A Two-Tier GDSF–GDS Caching Framework for High-Performance Content Delivery Networks.
Year: 2026
Type: article
Abstract: This paper presents a two-tier caching mechanism for content delivery networks (CDNs) that employs Greedy Dual Size Frequency (GDSF) in RAM and Greedy Dual Size (GDS) in SSDs. The proposed hybrid GDSF–GDS strategy is evaluated against widely used baseline policies, including LRU, LFU, and single-tier GDS-based caching. The design separates small, high-frequency objects in RAM from larger, low-frequency objects in SSDs to balance freshness, cost, and performance. Cost is modeled in terms of storage-tier utilization and object size awareness, capturing the trade-off between limited RAM resources and SSD capacity. Simulation experiments using 100,000 Zipf-distributed requests (α = 1.8) demonstrate a significant improvement in cache hit ratio compared to LRU, along with a substantial reduction in origin server fetches and lower access latency. A dynamic Time-to-Live (TTL) policy further maintains content freshness. Overall, the results indicate that the proposed two-tier GDSF–GDS approach improves efficiency and reduces latency in CDN environments.
Authors: Anwar A. Alhenshiri, Sahar E. Abodabous
Institution: Misurata University
Language: English
Topic: Caching and Content Delivery
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7153993183