A Survey of Edge Computing: Architecture, Technologies, and Research Challenges
Abstract
Edge computing has emerged as a key paradigm for addressing the limitations of centralized cloud computing in latency-sensitive, bandwidth-intensive, and geographically distributed applications. By bringing computation and storage closer to data sources and end users, edge computing enables new classes of applications in domains such as the Internet of Things (IoT), autonomous systems, augmented reality, and real-time analytics. This survey presents a comprehensive overview of edge computing, tracing its historical motivations, architectural models, enabling technologies, and application domains. We analyze the relationship between edge computing and related paradigms such as cloud computing and fog computing, and we discuss core challenges in resource management, security, programmability, and scalability. The paper concludes by identifying open research directions and future trends shaping the evolution of edge computing systems.
________________________________________
1. Introduction
The rapid proliferation of connected devices, sensors, and intelligent applications has fundamentally altered the landscape of distributed computing. Traditional cloud computing models, which centralize computation in large-scale data centers, have proven highly effective for elastic workloads and batch analytics [1]. However, emerging applications increasingly demand low latency, high reliability, real-time responsiveness, and efficient use of network bandwidth. These requirements expose inherent limitations in cloud-centric architectures, particularly when data sources are geographically dispersed or network connectivity is intermittent [2].
Edge computing has emerged as a response to these challenges by pushing computation, storage, and control closer to where data is generated and consumed [3]. This shift reduces end-to-end latency, alleviates core network congestion, and enables localized decision-making. As a result, edge computing has become a foundational technology for IoT ecosystems, smart cities, industrial automation, and next-generation mobile networks.
This survey provides a systems-oriented overview of edge computing, emphasizing architectural principles, enabling technologies, and open research challenges.
________________________________________
2. Background and Motivation
2.1 Limitations of Centralized Cloud Computing
While cloud computing offers scalability and elasticity, several structural limitations motivate edge computing adoption:
1. Latency Constraints: Applications such as autonomous driving, industrial control, and AR/VR require response times that centralized clouds cannot consistently meet due to network propagation delays [4].
2. Bandwidth Overheads: Continuous transmission of raw data streams (e.g., high-definition video) to the cloud is inefficient and costly [5].
3. Reliability and Availability: Cloud-centric designs assume stable connectivity, which may not hold in mobile, rural, or disaster-prone environments [6].
4. Privacy and Data Sovereignty: Regulatory frameworks such as GDPR restrict data movement across geographic boundaries, motivating local processing [7].
These constraints motivate architectures that distribute intelligence closer to the data source.
________________________________________
2.2 Evolution of Distributed Computing Paradigms
Edge computing builds upon prior paradigms in distributed systems research. Client–server computing introduced basic distribution, while content delivery networks (CDNs) demonstrated the benefits of placing resources closer to users [8]. Mobile computing explored computation offloading from constrained devices [9], and peer-to-peer systems emphasized decentralization at scale [10].
Edge computing integrates these ideas but distinguishes itself through tight coupling with network infrastructure and cloud orchestration frameworks.
________________________________________
3. Defining Edge Computing
3.1 Conceptual Definition
Edge computing refers to a distributed computing paradigm in which computation, storage, and networking resources are deployed closer to data sources and end users, often at the boundary of access networks [3]. Edge nodes may include gateways, base stations, routers, micro data centers, or even end devices.
Unlike traditional cloud computing, edge computing emphasizes locality, context-awareness, and real-time processing. Unlike purely decentralized systems, it typically operates in coordination with centralized cloud services.
________________________________________
3.2 Relationship to Fog and Cloud Computing
Fog computing, introduced by Cisco, describes a hierarchical model that extends cloud services across the continuum from the core to the edge [11]. In practice, fog and edge computing overlap significantly, with fog often emphasizing orchestration across multiple layers.
Cloud computing remains complementary. Modern systems increasingly adopt a cloud–edge continuum, in which workloads are dynamically placed based on latency, cost, and policy constraints [12].
________________________________________
4. Edge Computing Architectures
4.1 Hierarchical Models
Most edge computing systems adopt hierarchical architectures consisting of device, edge, and cloud layers [13]. Latency-sensitive tasks are processed near the edge, while compute-intensive analytics are delegated to the cloud.
________________________________________
4.2 Edge Nodes and Infrastructure
Edge nodes vary widely in capability, ranging from embedded systems to micro data centers deployed at cellular base stations or ISP aggregation points [14]. Compared to cloud data centers, edge nodes are resource-constrained and geographically distributed, complicating management and scheduling.
________________________________________
4.3 Virtualization and Containers
Lightweight virtualization technologies such as containers are widely used at the edge due to their low overhead and fast startup times [15]. However, existing orchestration systems such as Kubernetes require adaptation to handle intermittent connectivity, node heterogeneity, and limited resources [16].
________________________________________
5. Networking and Communication
5.1 Mobile Edge Computing and 5G
Edge computing is closely integrated with next-generation mobile networks. Mobile Edge Computing (MEC) places compute resources within cellular infrastructure to support ultra-low-latency applications [17]. 5G introduces features such as network slicing and ultra-reliable low-latency communication (URLLC), which are critical for edge workloads.
________________________________________
5.2 Data Placement and Mobility
Efficient data placement is a central challenge in edge systems. Decisions must consider latency, bandwidth, energy, and mobility patterns [18]. User and device mobility further complicate state consistency and task migration.
________________________________________
6. Application Domains
6.1 Internet of Things
IoT is a primary driver of edge computing. Edge processing enables real-time analytics, filtering, and anomaly detection close to sensors, reducing cloud load and improving responsiveness [19].
________________________________________
6.2 Autonomous and Cyber-Physical Systems
Autonomous vehicles, drones, and robotic systems rely on fast perception–action loops. Edge computing enables cooperative perception and localized coordination while leveraging the cloud for training and global optimization [20].
________________________________________
6.3 AR/VR and Interactive Media
AR and VR applications demand extremely low motion-to-photon latency. Edge rendering and caching reduce end-to-end delays and improve user experience [21].
________________________________________
6.4 Industrial and Smart Infrastructure
In industrial automation and smart grids, edge computing supports real-time control, fault detection, and predictive maintenance under strict reliability constraints [22].
________________________________________
7. Resource Management and Scheduling
7.1 Heterogeneity
Edge environments are highly heterogeneous in terms of hardware, power budgets, and connectivity. Resource management frameworks must account for diverse CPU architectures, accelerators, and energy constraints [23].
________________________________________
7.2 Task Offloading
Task offloading decisions involve tradeoffs between latency, energy consumption, and resource utilization. Research approaches range from heuristic algorithms to optimization models and machine learning-based policies [24].
________________________________________
7.3 Scalability and Elasticity
Unlike cloud environments, edge resources cannot be easily overprovisioned. Scalability relies on workload migration, replication, and cooperation among nearby nodes [25].
________________________________________
8. Security and Privacy
8.1 Threat Model Expansion
Deploying compute resources in physically accessible locations increases exposure to attacks such as tampering, malware injection, and side-channel attacks [26].
________________________________________
8.2 Privacy Preservation
Local processing can reduce data exposure, but trust management remains challenging. Trusted execution environments and secure enclaves are promising solutions [27].
________________________________________
8.3 Secure Orchestration
Coordinating workloads across cloud and edge environments requires secure control planes, authentication, and authorization mechanisms that operate under partial connectivity [28].
________________________________________
9. Programming Models and Systems Support
9.1 Edge-Aware Abstractions
Edge computing requires new programming abstractions that expose locality, mobility, and resource constraints [29].
________________________________________
9.2 Serverless at the Edge
Serverless computing has been extended to edge environments to support event-driven execution [30]. However, cold-start latency, state management, and debugging remain open challenges.
________________________________________
9.3 Observability
Monitoring and debugging edge systems is difficult due to scale, heterogeneity, and intermittent connectivity. Improved observability tools are essential for reliable operation [31].
________________________________________
10. Open Research Challenges and Future Directions
Key open challenges include energy-efficient edge architectures, AI-driven resource management, interoperability across vendors, and integration with emerging accelerators. Future systems are likely to adopt autonomous management frameworks that continuously optimize workload placement across the cloud–edge continuum [32].
________________________________________
11. Conclusion
Edge computing represents a fundamental shift in distributed systems design, addressing the limitations of centralized cloud architectures by moving computation closer to data sources and users. While enabling powerful new applications, edge computing introduces significant challenges in resource management, security, and programmability. Continued research is essential to fully realize its potential and to integrate edge computing into a unified global computing fabric.
________________________________________
References
[1] P. Mell and T. Grance, “The NIST Definition of Cloud Computing,” NIST, 2011.
[2] L. Barroso, J. Clidaras, and U. Hölzle, The Datacenter as a Computer, Morgan & Claypool, 2013.
[3] W. Shi et al., “Edge Computing: Vision and Challenges,” IEEE Internet of Things Journal, 2016.
[4] M. Satyanarayanan, “The Emergence of Edge Computing,” IEEE Computer, 2017.
[5] A. Botta et al., “Integration of Cloud Computing and Internet of Things,” Future Generation Computer Systems, 2016.
[6] E. Cuervo et al., “MAUI: Making Smartphones Last Longer,” MobiSys, 2010.
[7] European Union, “General Data Protection Regulation,” 2016.
[8] B. Krishnamurthy et al., “On the Use and Performance of Content Distribution Networks,” IMW, 2001.
[9] S. Abolfazli et al., “Cloudlets: At the Edge of the Internet,” Mobile Cloud Computing, 2012.
[10] S. Androutsellis-Theotokis et al., “A Survey of Peer-to-Peer Content Distribution,” ACM Computing Surveys, 2004.
[11] F. Bonomi et al., “Fog Computing and Its Role in the Internet of Things,” MCC, 2012.
[12] K. Ha et al., “Towards Wearable Cognitive Assistance,” MobiSys, 2014.
[13] Y. Mao, C. You, J. Zhang, K. Huang, and K. Letaief, “A Survey on Mobile Edge Computing,” IEEE Communications Surveys & Tutorials, 2017.
[14] M. Chiang and T. Zhang, “Fog and IoT: An Overview,” IEEE Internet of Things Journal, 2016.
[15] D. Merkel, “Docker: Lightweight Linux Containers,” Linux Journal, 2014.
[16] P. Morabito et al., “Orchestrating Containers at the Edge,” IEEE Communications Magazine, 2018.
[17] ETSI, “Mobile Edge Computing (MEC),” ETSI White Paper, 2014.
[18] S. Wang et al., “Dynamic Service Placement for Mobile Micro-Clouds,” MobiSys, 2015.
[19] L. Atzori, A. Iera, and G. Morabito, “The Internet of Things,” Computer Networks, 2010.
[20] J. Liu et al., “Edge Computing for Autonomous Driving,” IEEE Network, 2018.
[21] X. Ran et al., “DeepDecision: A Mobile Deep Learning Framework,” INFOCOM, 2018.
[22] R. Buyya et al., “Industrial IoT and Edge Computing,” Future Generation Computer Systems, 2018.
[23] T. Taleb et al., “On Multi-Access Edge Computing,” IEEE Communications Magazine, 2017.
[24] Y. Chen et al., “Efficient Offloading for Mobile Edge Computing,” IEEE Transactions on Vehicular Technology, 2016.
[25] S. Yi et al., “A Survey of Fog Computing,” ACM Computing Surveys, 2017.
[26] R. Roman et al., “Securing the Internet of Things,” Computer, 2013.
[27] V. Costan and S. Devadas, “Intel SGX Explained,” IACR Cryptology ePrint Archive, 2016.
[28] A. G. Ranjan et al., “Security and Privacy in Edge Computing,” IEEE Internet Computing, 2019.
[29] M. Satyanarayanan et al., “Edge Analytics in the Internet of Things,” IEEE Pervasive Computing, 2015.
[30] L. Wang et al., “Peeking Behind the Curtains of Serverless Platforms,” USENIX ATC, 2018.
[31] P. Jamshidi et al., “Self-Learning Cloud Controllers,” ICSE, 2017.
[32] S. Dustdar et al., “Edge Intelligence,” Proceedings of the IEEE, 2019.