Managing the IoT: Edge Computing and SDN
By itself, connecting a lot of ‘things’ to the Internet will not create a broadly useful Internet of Things (IoT) — the kind of IoT that will transform everything from factories to Smart Cities. That kind of IoT requires massive available computing power and networks with the capacity for staggering numbers of endpoint connections. Where (and at what cost) will you perform that computing? And how do you manage networks that connect hundreds of thousands of things? The answers are edge computing and Software-Defined Networking (SDN).
Understanding these two concepts requires a basic description of the four layers of the IoT:
- Sensing and control layer comprising ‘things’ (endpoint devices) that serve as data sources and/or perform some action such as a sensor and include SDN control logic
- Physical network layer that is used mostly for data backhaul
- Platform layer to manage connectivity and Operations and Maintenance (O&M)
- Application layer for data analysis and applications control
Edge Computing Advantages
The advent of cloud computing posits that management, analysis, control, and data processing tasks are best performed in data centers that host the platform and application layers.
Performing all computing tasks in data centers has drawbacks, especially for industrial applications that require real-time performance. IoT architects have proposed that an open platform at the network edge will reliably perform tasks such as connection, computing, storage, and application installation. This edge-computing platform is close to the things that sense conditions and control actions. Considering the location, the edge-computing platform is generally implemented in an IoT gateway, where most IoT data will be aggregated in the foreseeable future.
Edge computing offers four primary advantages over traditional networking:
- Edge computing meets the needs of applications for real-time high performance. Delays in response time for control functions processed in the cloud will often be too long; therefore, some classes of analysis and control functions must be implemented at the network edge to meet specific, real-time service requirements. For example, in production control, the maximum delay in service control is often 10 ms or less. For automated driving, control delays must be within several milliseconds.
- Due to increases in local storage capacity, edge computing is well-suited to handle data adaptation and aggregation tasks. This approach is useful for sensing and control layers that involve the unification of complex, heterogeneous communications technologies and data protocols.
- If bulk IoT data were sent from edges to data centers for processing, the cost of network operations would be unnecessarily high. For instance, temperature sensors need only report abnormal changes to the data center. Likewise, in the realm of facial recognition, rather than sending raw image data, only a few key characteristics need to be uploaded to the data center.
- Edge computing is reliable. Data center processing adds a level of complexity that inherently increases risk in many industrial applications. For maximum reliability, edge systems must maintain a certain level of autonomy; for manufacturing control systems, the collaboration between distributed intelligence and autonomous systems allows network edges to help secure the survivability of individual nodes and the entire system. Even with a basic system like connected streetlights, local controllability ensures pedestrian and traffic safety in the event that the city data center goes offline.
- For many manufacturing systems, access network security is especially important. For the IoT, the network connecting the sensing layer with the data platform layer is usually the most vulnerable. To overcome this exposure, hardened security is best performed at the IoT gateways nearest the network edge.
Opening up Edge Computing
As IoT technology becomes more widely employed, edge computing turns out to be a prerequisite for implementing many industrial applications. To help define edge computing architectures, open standards, and rapid adoption of IoT ecosystems, Huawei has joined the China Academy of Information and Communications Technology, Shenyang Institute of Automation, Intel, ARM, and iSoftStone to form the Edge Computing Consortium.
In support of this trend, Huawei has used edge computing as part of its Connected City Lighting and Connected Elevator solutions since early 2016.
IoT and Edge Computing Management Challenges
By 2025, Huawei projects that the application of ICT technologies will create 100 billion connections, 90 percent of which will come from various intelligent edge applications that connect things to things. This vast number of connections poses a great challenge to the management, maintenance, and control of the IoT.
Consider an ‘Energy Internet’ IoT project Huawei delivered in Nigeria that involved deploying 300,000 electric meters and communications modules, and tens of thousands of IoT gateways. Traditional network management systems were unable to handle such huge numbers of addressable components.
In this project, edge computing provided the advantages described earlier but also imposed new requirements on the management system, which in the past was responsible simply for network facilities. Now, the system must also manage the computing and storage resources of IoT gateways, and open-source third-party applications running on those gateways.
While the IoT and edge computing create exciting opportunities to use real-time data in countless scenarios, manually managing the edge network will grow increasingly impractical. The numbers and types of IoT devices are proliferating, networks are exposed to security threats, and management complexity is rising. The explosion of data collected and conveyed through IoT networks frequently requires near-immediate response times for safety, security, and production continuity.
Enabling IoT Management through SDN
SDN can cost-effectively virtualize IoT networks to provide automatic device reconfiguration and bandwidth allocation to boost performance and conserve bandwidth. SDN simplifies network management for even the most complex networks by offering plug-and-play device setup and deployment. SDN ensures security by detecting and resolving threats through automated application of security policies and improved access control with the benefit of greater traffic transparency at the network’s edge.
Enabled by the policy-driven management of massive numbers of data center and edge computing devices, SDN supports the unified management of all ICT resources, including the unification of lifecycle management and the control of virtual machines, containers, and their mirror files.
Southbound controller interfaces enable centralized management of sensors, terminals, communication modules, IoT gateways, and other devices. Plug-and-play technology simplifies the management of multiple devices by implementing automatic deployment, security authentication, status monitoring, and remote upgrades. In the future, SDN is expected to leverage Artificial Intelligence (AI) to implement in-depth fault analysis and automated troubleshooting.
The Role of Data Management
SDN also helps manage applications, including the ability to handle data subscription and distribution. This is especially useful when edge computing is implemented on open platforms that support third-party applications and edge services while decoupling network and data connections. SDN supports quick and flexible interworking with multiple data platforms through unified management of the delivery, installation, operation, and deletion of third-party applications.
The need for data management is a major difference between IoT management and traditional network management. Traditional connectivity technologies lack the depth of in-service analysis that is available in the current generation of network monitoring applications. For the IoT, collecting, analyzing, and uploading data is the point — and SDN provides the network flexibility to include the widest range of scenarios.
SDN and IoT Control
Support for data uploading is another important SDN feature. For example, agent software on IoT gateways interacts with application platforms in data centers or with IoT platforms. This approach is supported by Huawei’s OceanConnect, an open ecosystem built on IoT, cloud computing, and Big Data technologies, and GE’s Predix, a cloud-based industrial Internet platform. Data can be sent to controllers for simple protocol processing before being forwarded to application platforms or IoT platforms. Network and data connections are decoupled from each other, with controllers performing data distribution and subscription and unified data uploading.
Integration of northbound and southbound interfaces is also helpful for IoT applications. The IoT is an ecosystem involving different applications and domains for which no single vendor can supply a complete end-to-end IoT system. By providing northbound and southbound interfaces that support unified standards for quick integration with third-party systems, SDN can simplify the configuration of IoT platforms.
Wide-Area Network Support
SDN removes the restrictions of traditional network architectures and is proving useful in Wide-Area Networks (WANs), with the goal of adapting networks to deal with rapid changes in industry demands. WAN controllers implement unified and centralized network management and global business model abstractions. Standard northbound interfaces for upper layer applications enable fast, inter-application integrations and day-to-day adjustments for SDN-equipped business networks.
Huawei SDN/IoT Solutions
With all these functions, the use of SDN technology, especially controllers, will resolve issues such as IoT management, control, maintenance, and interoperability. Since launching an agile IoT solution at the 2015 Huawei Network Congress, Huawei has deployed systems in a variety of domains, including electrical power transmission, connected lighting systems, intelligent buildings, and the Internet of Vehicles in collaboration with partners and other mainstream vendors.
Open Platforms and Simplified SDN Management
IoT platforms are much more complex than the simple connection of edge devices to the network and will always involve many different business applications and technical considerations. As the number of devices continues to grow (from an increasing number of vendors), networks will become more complex, harder to manage, and more vulnerable to security threats.
Fortunately, the combination of edge computing and SDN offers a path forward through many of the issues and complexities that the IoT services industry is facing. With edge computing on open platforms and simplified SDN management, IoT implementations will feature reliable, real-time, secure operations capable of taking advantage of the growing range of connected ‘things’ and the valuable data they generate.