Today's engineers are trained and prepared to tackle the challenges, but the IoT just may not happen as quickly as the pundits say.
Peel away the hype surrounding the Internet of Things, and beneath it all is a veritable rat's nest of technical challenges that must be tackled before its full realization.
More than ever before, there will be huge demands on nodes and the network infrastructure, creating significant engineering challenges. The IoT will require engineers to look at end-to-end network solutions that extend far beyond that of the datacenter. Along every stage of the network, engineers will need to make efficient use of silicon that meets both performance and budgetary requirements.
Engineering an IoT will require a deep appreciation of the scale of the data-capture devices and the diverse locations transmitting information to the datacenter. It is now accepted that IoT will be one of the key drivers for the big-data phenomenon, but the actual means of data transfer at both the physical and application layers is rarely a key consideration.
Other ugly technical challenges include the last mile; the availability of suitable networking technologies that scale to tens of millions of devices; the significant increase in basestation processing and storage requirements; and the development of new technologies for efficient interconnect and workload processing at the datacenter.
Networking has for decades faced challenges providing a high QoS at the last mile. Traditionally the Internet's last mile has been homes and small offices, which are both stable locations that require primarily downstream bandwidth. But over the past five years the last mile has extended to include smartphones and tablets, increasing the need today for upstream bandwidth.
Wireless cellular networking standards such as Long Term Evolution (LTE) have been established to increase bandwidth and reduce latency to the end user. While deployment of LTE (or "4G") networks may be patchy, they serve as an indicator of just how network requirements are changing in order to provide higher levels of QoS.
Further, a single solution will not be able to service the IoT needs of a vast array of sensor types and network characteristics, while at the same time also achieve optimal performance and use of resources. When it comes to meeting the needs of all stakeholders in an IoT project, an engineer will not be able to rely simply on the network hardware to do all the heavy lifting.
That is especially true when making modifications to the underlying network is not feasible, in which case an overlay network needs to be created. Such a network is typically defined at the application layer with heuristics that can deliver a QoS that meets users' requirements.
Network overlay design typically involves working with the underlying network to engineer application-layer protocols that meet design goals. In the past this has included the application-layer implementation of IP-layer protocols such as Anycast and the creation of swarming algorithms in distributed networks such as Bittorrent. Though it may sound simple, it is diabolically complex to design and implement an overlay that meets the performance requirements of the rest of the network without the luxury of starting from scratch.
The IoT should not be viewed as a network with one-way traffic generated by sensors that upload data to the cloud. Giving users the functionality to poll areas of the network gives rise to the need for protocols like Anycast, which calls for queries to be routed to a sensor that meets specific criteria, such as location and type. Such network protocols will turn IoT sensors into bi-directional nodes that users can home in on after analyzing the mountain of data generated by millions of sensors.
Engineers also will need to design and deploy an IP network with a lightweight protocol that is suited to the types of data collected. The lightweight protocol has to take into account the scale of the IoT while at the same time handling data integrity and transmission problems posed by the remote location of sensors. Designing in techniques for error checking and packet retransmission won't be an option.
The need for lightweight network and authentication protocols arises because millions of sensors have to be energy efficient. Practically, they will need to go into a low-power sleep mode, disabling most of the connectivity in order to minimize power draw.
Engineers should be mindful that, not only will lightweight network protocols need to be considered, but data security requirements will dictate the implementation of lightweight authentication protocols. These protocols need to accommodate the compute power that will be available on low-cost, widely deployed sensors while maintaining the QoS level required by the user.
Moreover, the basic QoS that a typical IoT device will demand is higher than anyone could have conceived of. The need for integrity, whether it be the correct reporting of data or the secure transmission of data from source to datacenter, will be vital to the creation of a true IoT and will place the need for network engineering at the heart of designing it.
The good news in all of this is that today's engineers are trained and prepared to tackle these challenges -- the IoT just may not happen as quickly as the pundits say.
With its Embedded G-Series Family of processors including APUs, CPUs and SOCs. AMD is targeting low-power, low-cost, low-maintenance communications and infrastructure.
The AMD Embedded G-Series Family of processors – including APUs and SoCs – are designed to meet the growing demands of communications and infrastructure applications. With a roadmap that includes both ARM and x86 based processors the G-Series offers high performance, low-power, low-cost and low-maintenance solutions. The AMD G-Series Family of processors provide the foundation to build a network that can support the demands generated by an Internet of Things.
Read more about the challenges of IoT here:
— Lawrence Latif is a Technical Communications Manager at AMD. He has published peer-reviewed research and has over a decade of experience in enterprise IT, networking, system administration, software infrastructure, and data analytics. Lawrence holds a BSc in computer science and management from Kings College London, an MSc in systems engineering management, and a PhD in electronic and electrical engineering from University College London.