Solving the IoT Network Puzzle
May 7, 2014, Internet of Things Conference, Santa Clara, CA—Tareq Bustami from Freescale Semiconductor described the convergence of networking technologies for IoT. The changes in architectures and technologies will allow data to move across many platforms and communications networks with minimal friction.
The IoT can be shown to be associated with many network concepts from basic protocols to software defined networks (SDN) and network function virtualization (NFV). The underlying concept for the communications is that of redistributed intelligence. The challenge is that network infrastructure changes slowly, while the IoT endpoints are emerging before the networks are ready for them.
The SDN is a response to the need for increased programmability and centralized control. The architectures help to scale IP and Ethernet networks into more adaptive systems and help to deploy apps into the Internet by separating the app and data planes. The increased automation enables services to create, install, deploy, monitor, and analyze all network operations.
NFV moves network services into software-defined services. Now, the technology is moving into standard SoCs and virtualized software. The primary users are service providers for the mobile core and network functions like broadband remote access servers. This technology is not for real-time functions or any functions that is latency sensitive.
By redistributing intelligence and moving the intelligence to the context, users get better coordination of their apps and data. These characteristics are enhanced by the future technologies which will evolve the network. For example, the 5G infrastructure will have a new architecture with widely distributed intelligence.
These architectural changes to put the intelligence where it is needed is ideal for the IoT. Sensors can communicate to a gateway, which provides connectivity to the access structures of a network and to the cloud. For example, a smart car would have different levels of intelligence depending on the functions and connections to adjacent vehicles and other infrastructure.
A cloud radio access network like 5 G would provide access and GbE integration to 100 times the number of users of current systems and would give all those users up to 100 times the bandwidth. These and other data sources need the capabilities of NFV.
The challenges for these technologies include the need for new business models within the next 3-5 years. The new systems will have to be integrated and also co-exist with existing base systems driving the need for new ecosystems.
The parallel move to multicore processors will provide greater ease of use, the ability to scale, and offer better software support. Open instruction sets will permit peripherals to be considered as software objects. Some version of C running on these processors will enable programmable packet engines that raise the levels of abstraction for management and configuration functions.
The open instruction sets will also enable simpler scaling, since more software will be able to access the differing capabilities in the platforms. The platforms themselves will be more easily reused creating economies of scale for the semiconductor manufacturers. The software will be written to take advantage of vendor-neutral APIs, larger and more capable libraries of functions, and other services and resources.
