Tao Zhang, Visiting Chair Professor, National Chiao Tung University and T. Russell Hsing, Chair Professor, National Chiao Tung University
Over the past decade, cloud computing has played a dominant role in supporting the applications we rely on today. Networks have served mainly as communication pipes connecting users to the cloud and with each other. This cloud plus communication pipe model, however, is no longer adequate for supporting the many emerging applications such as the Internet of Things (IoT), next-generation mobile networks, vehicular networks, automated manufacturing, smart cities, drones, smart grids, e-health, Virtual Reality (VR), and Augmented Reality (AR). For example, many IoT applications cannot tolerate the delays incurred by cloud computing. Client devices are creating a vast and ever-growing amount of data that needs to be processed locally because sending all the data to the cloud will often be infeasible due to network bandwidth constraints and regulatory restrictions.
Connecting every device directly to the cloud can often prove impractical due to limited resources on the devices, the excessive complexity and cost to add cloud connectivity and management to all IoT devices, and scalability limitations. Many resource-constrained devices will also require local services to help perform many tasks that they cannot carry out alone. Such tasks can range from computational-intensive user applications to security protection measures that require heavy processing or global information that the devices do not have. Future networks will also need to better integrate with new computing capabilities, especially around the end users. Such computing and networking convergence can enable a wide range of new capabilities such as time-sensitive and on-demand intelligent networking at the edge and among client devices, dynamically adjusting radio channel coding in response to changing user demands and communication environments. This convergence of computing and networking will also enable ambient intelligence allowing networks, client devices, things, applications, and data to integrate more closely around the user.
To meet the needs of the emerging applications and networks, the clouds are descending toward the ground and even dispersed among the client devices – forming fog computing and networking or fog.
Fog envisions open and standards-based horizontal architectures for distributing functions (from computing to storage to control and to networking functions) closer to users, not just to any specific type of network edge device but anywhere along the cloud-to-thing continuum that can best meet user requirements. Fog will integrate with the cloud to enable seamless computing throughout the cloud-to-thing continuum. Computing functions can be deployed and subsequently moved anywhere along this computing continuum. Computing resources along the continuum can be pooled together to support user applications. Users will not have to differentiate whether their services and applications come from the cloud or the fog – the services will run where they can best meet user requirements. Fogs will work autonomously when connectivity to the cloud is unavailable. They can also collaborate with each other to carry out tasks for the users.
Fog empowers and extends the cloud. Fog can help connect a vast range of endpoints to the cloud more practically by using simpler local procedures and protocols to interact with the endpoints so that these endpoints can remain simple, “dumb”, and low cost. Fog can also act as the proxy of the cloud to bring cloud services closer to the endpoints. For example, fog nodes inside a radio access network can act as the proxy of the cloud to deliver cloud services to mobile users.
Fog will enable many new services that the cloud alone cannot effectively support. For example, a fog system, more powerful than the endpoints, can provide local security services to protect the endpoints. Such security services may include local monitoring, security credential and software updates, and malware detection and protection for the endpoints. Fog-as-a-Service will allow users to access private and public fog systems and services deployed close to them. Users can store their data and host their applications in these nearby fog systems. Users may also rent storage spaces or computing servers from these fog systems and manage their data and applications by themselves. Users may even receive turnkey services from a Fog-as-a-Service operator.
Achieving the full potential of fog computing and networking creates a futile ground for research and innovation. It calls for rethinking of computing and networking architectures. What should be the roles of client and network edge devices that are becoming ever more powerful and prevalent? How should future networks better integrate with the emerging computing capabilities such as AI and machine learning? What will the fog architectures look like? How best to distribute, orchestrate, manage resources in distributed fog networks? How should fog integrate with the cloud, with radio access networks, with other edge networks, and with client devices? How to secure a large distributed fog system that often needs to operate in vulnerable but under-managed environments? The list goes on.
Answers to these challenges will bring profound disruptions to technology and businesses over the next decade – the decade of fog computing.
Learn more at Fog World Congress, the first conference that brings industry and research together to explore the technologies, challenges, industry deployments and opportunities in fog computing and networking. Scheduled for Oct. 1-3 in San Francisco, Fog World Congress will feature presenters, panels and topics that focus on the transformational capabilities of fog computing in IoT, 5G and AI.