CXL Powered Memory Intensive Graph Processing: A Comprehensive Survey

Graph processing faces severe memory bottlenecks as data scales exponentially, with traditional methods struggling to handle graphs containing billions of vertices and edges. This survey investigates how Compute Express Link (CXL) technology addresses these limitations in large-scale graph analytics. We systematically analyze memory challenges across core algorithms—including k-edge/k-core decomposition, strongly connected components, and knowledge graph embedding search and demonstrate how CXL provides transformative solutions through expanded memory capacity, tiered memory management, and shared memory support. These mechanisms enable in-memory processing of previously disk-bound graphs, intelligent hot-cold data placement, and efficient parallel processing with reduced communication overhead. However, realizing CXL’s potential requires addressing critical challenges in data placement optimization, remote memory access latency, and concurrent access control. This work provides the first systematic survey for understanding CXL’s role in next-generation graph processing systems, bridging the gap between emerging memory technologies and graph analytics requirements. Our analysis offers valuable insights for researchers and practitioners seeking to leverage CXL for memory-intensive graph applications.