2026· IEEE Transactions on Networking· Vol 34, pp. 6045-6060· 0 citations· 48 references
Computer Science
Abstract
The rapid expansion of cloud computing, big data analytics, and artificial intelligence has positioned data centers as the backbone infrastructure of modern computing. Data center networks (DCNs) play a critical role in determining overall system performance and reliability. Existing DCN architectures face limitations, such as difficulties in balancing throughput and latency, insufficient fault-tolerance, and high expansion costs. To address these challenges, we propose ACDC (Augmented Cube-based Data Center), a novel server-centric DCN topology based on augmented cubes that achieves superior performance while maintaining cost-effectiveness through exclusive use of dual-port servers and low-port commodity switches. Firstly, we analyze the key features and properties of ACDC, with a focus on scalability and network diameter, establishing rigorous theoretical foundations for the proposed architecture. Secondly, we present comprehensive routing algorithms including ARouting for fault-free scenarios and AFR for fault-tolerant communication. Finally, extensive experimental evaluations demonstrate ACDC’s superior performance compared to state-of-the-art DCN architectures. Experimental results show that ACDC achieves a network diameter approximately 75% smaller than HSDC and 50% smaller than AQDN. Furthermore, ACDC maintains comparable throughput to the Fat-Tree under random traffic scenarios, while demonstrating substantial advantages under high-density all-to-all communication patterns, achieving at least 69.1% improvement in average throughput and at least 40.7% reduction in flow completion time compared to AQDN, DCell and FiConn. These confirm that ACDC strikes a good balance among performance, cost-efficiency, scalability, and fault-tolerance in contrast to the state-of-the-art DCN architectures.
The evaluation of the proposed architecture through analytical models and simulation-based evaluations shows that the proposed architecture can reduce the latency onto 65 percent of the time relative to the conventional cloud-based architecture, affirm the claim that edge computing is an essential enabler of the next-generation applications that demand deterministic response time, high reliability and localized intelligence.
Priya Natarajan· International Journal of Mod...· 0 citations
The study concludes that intelligent edge computing architectures will play a vital role in supporting future real-time applications and next-generation 6G-enabled digital ecosystems.
Alan Bundy· International Journal of Mod...· 0 citations
Performance evaluation is essential for understanding, comparing, and improving computing systems, including Distributed Computing Continuum Systems (DCCS). In recent years, computational requirements have changed substantially with the growth of artificial intelligence and large-scale data-driven applications. These application tasks are increasingly distributed between resource-intensive data centers and resource-constrained edge environments. In this context, novel computing continuum architectures and algorithms are emerging, creating a need for transparent and consistent performance evaluation. However, existing evaluation practices often focus on isolated dimensions, such as computation, networking, energy efficiency, or application-level quality, and therefore provide only a partial view of cross-layer DCCS behavior. This paper presents a structured taxonomy of performance metrics for DCCS. The taxonomy organizes metrics into computing-level, network-level, and application/user-level categories, while also highlighting emerging dimensions such as sustainability, observability, adaptability, data locality, migration awareness, and continuum fragmentation. Further, we provide mathematical formulations and discuss their relevance to heterogeneous and dynamic continuum environments. We also summarize metric acquisition requirements in terms of acquisition scope, acquisition phase, and measurement method. These requirements help clarify whether a metric can be collected from a single node, multiple nodes, or the full system, and whether it is more suitable for operational monitoring or experimental evaluation.
Praveen Kumar Donta, Boris Sedlak, Alfreds Lapkovskis et al.· arXiv.org· 0 citations
A structured review of optimization models in cloud and data center environments using a Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guided methodology covering literature from 2016 to 2025 reveals that the adaptive methods can improve throughput, reduce latency, and enhance energy efficiency under specific datasets, simulation settings, traffic models, and network configurations.
S. Alanazi· Journal of Advances in Infor...· 0 citations
Lots of web based cloud computing applications like high performance computing, content delivery and video streaming, enterprise applications, artificial intelligence and machine learning require large computing power and memory resources which can only be provisioned by a Data Center Network (DCN). A DCN is a collection of a large number of high performance servers and network devices interconnected to achieve a specific task or application. Now a days there is a paradigm shift to replace many electric network devices interconnecting high end servers with a single large port count optical interconnect [1]. The port count and blocking probability of a port of optical interconnects in DCN are important parameters as they affect the overall latency, complexity of routing, and energy consumption in the network; a lower port count of interconnects means a large hop count of the overall network. If hop count is large then at each hop there is store and forwards mechanism which increases latency, complexity of routing and energy consumption all of which are negative markers of performance. Likewise high blocking probability refers to more re-transmissions of packets. There is a growing demand for DCNs, due to compute-hungry demand of modern web-based applications like video streaming, social networking, powerful search engines, and email repositories, just to name a few. Ever since the invention of network packet switching devices, there is a quest for large port count network switches, which actually reduces the number of network switching devices in the network. To enhance the performance of such applications we need optical interconnects which have large port count. The contributions of this paper are threefold i. The design considerations associated with large port count optical interconnects are identified, ii. Analysis of existing architectures in terms of port count and optical components affecting port count, iii. A novel large port count architecture called Enhanced Port Count (EPC) is proposed that provides port count benefit over existing architecture.
The proliferation of data-intensive applications, such as generative AI, has substantially increased the demand for low-latency and energy-efficient data center networks (DCNs). Supporting these applications requires large-scale deployment of servers and GPUs, which in turn necessitates highly scalable DCNs. To meet these performance and scalability requirements, hierarchical optical DCNs leveraging optical circuit switches (OCSs) have emerged as a promising approach. The reliable operation of hierarchical OCS-based DCNs depends on accurate fiber-link topology information. However, topology discovery at the fiber layer is inherently challenging because OCSs operate transparently without signal processing or optical power monitoring capabilities. Although prior studies have investigated fiber-layer topology inspection methods for OCS-based DCNs, their applicability is limited to conventional duplex fiber-pair deployments and does not extend to (i) deployments employing bidirectional (Bi-Di) transmission over a single fiber core or (ii) deployments requiring per-core fiber management due to OCS constraints. To bridge this gap, we propose a fiber-link discovery (FLD) algorithm that correctly identifies unidirectional fiber topology in hierarchical OCS-based DCNs, requiring only O(h log2 L) discovery steps, where h is the number of stages and L is the number of fibers between each pair of adjacent-stage OCSs. Simulation results demonstrate that our algorithm achieves up to 341.3× faster topology identification than the baseline method while guaranteeing correctness.
Kazuya Anazawa, Toru Mano, Yoshiaki Sone et al.· International Conference on...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
