Published November 16, 2026 | Version v1

Towards a New Model of Real-Time Graph Computation on Social Media

Description

Real-time analysis of social media data is increasingly constrained by the scalability, memory efficiency, and latency limitations of existing graph-processing frameworks, which exhibit different trade-offs in meeting real-time requirements. This study adopts a conceptual, literature-based comparative approach to examine four representative frameworks: Apache Hadoop with Giraph, Apache Spark with GraphX, Neo4j, and Apache Flink with Gelly, focusing on scalability, real-time processing, and memory management. Based on the identified limitations, a conceptual architecture is proposed that integrates dynamic graph partitioning, in-memory processing, and distributed memory management to support real-time graph computation in Social Media Networks (SMNs). The primary contribution lies in coordinating these established techniques within a unified processing pipeline designed to address the scalability and latency demands of social media graphs. The proposed architecture remains at the conceptual design stage and has not undergone direct distributed benchmarking. Nevertheless, implementation patterns from two related social-media graph studies provide an indicative, prototype-informed evaluation envelope, suggesting approximately 300–400 nodes, 1,100–1,400 edges, 150–220 KB of graph-memory context, sub-second average API response times, and graph recomputation times of approximately 50–330 seconds under study-scale conditions. These figures are estimates for evaluation planning rather than measured results or controlled comparisons among the four frameworks. Accordingly, the study establishes a structured comparative foundation and identifies a staged validation process involving prototype development and performance evaluation using public social-network datasets. Its principal contribution is a technically motivated architectural design that informs future empirical investigation while explicitly acknowledging the absence of direct experimental validation and controlled cross-framework performance measurements.

Files

65.pdf

Files (840.3 kB)

Name Size
md5:3293b07e4aedd390373e186f620a72bf
840.3 kB Preview Download

Additional details

References