Bench Talk

(Source: mast3r - stock.adobe.com)

The trend toward cloud computing, communication services, and containerization has been well underway for years. This has led to a buildout of nearly 65 million square feet of data center space globally, with nearly 5 million square feet more currently under construction.¹ These data centers are typically extensive facilities located outside cities, where real estate prices and utilities lead to more favorable capital and operational expense figures. This model of building large data centers fits well with the internet revolution. However, the connectivity revolution—primarily driven by the 5G rollout, the explosion of the Internet of Things (IoT), and advances in augmented reality/virtual reality (AR/VR) wireless technologies—is leading to a different type of data center expansion (Figure 1). An expansion that is far less centralized and ultimately requires the placement of edge computing and data center services as near as possible to the end-user devices and autonomous systems that need bleeding edge latencies and throughput.

This blog discusses how the rise of IoT and 5G rollout is impacting edge computing and data center requirements to support these new ultra-reliable low latency communications (URLLC), massive machine-type communications (mMTC), and enhanced mobile broadband (eMBB) services.

Figure 1: An augmented reality application assists a technician in a smart factory setting, highlighting the convergence of IoT and AR technologies that demand ultra-reliable, low-latency networking solutions. (Source: zapp2photo - stock.adobe.com)

Traditional internet and cloud services are based on a model where massive amounts of data can be stored and processed from multitudes of end devices around the globe. These cloud services are now often containerized and supported by arrays of servers that can flexibly meet the demands of various applications with reasonable latencies. These latencies are usually in the tens of milliseconds, which is acceptable for most legacy applications, such as gaming, internet services, video streaming, audio streaming, storage, and high-performance computing. To reduce latencies for users, large cloud service companies will often distribute their most popular content to various data center locations within a reasonable distance of the users requesting that specific data. This is typically done with well-developed algorithms. However, with latency for these services still in the tens of milliseconds, it is well beyond the acceptable threshold for many applications, like modern autonomous vehicles, industrial IoT (IIoT), autonomous robotics, and edge processing.

Though 5G rollout is still in the early stages and most 5G services are still in the sub-6GHz (FR1) portion of the spectrum, these services are still rapidly setting the foundation for a huge shift in where and how data will be generated, processed, and distributed through networks. Many 5G URLLC applications are striving for less than one-millisecond latency, which is over a factor of 10 reduction in latency. Moreover, this reduction in latency is also paired with eMBB goals of achieving a 20-factor increase in mobile data speeds to an average of nearly 600Mbps, compared to the current but growing figure of average speeds around 30Mbps. Hence, the same networking infrastructure that supports sub-millisecond latencies will also need to support many times the throughput of current mobile broadband services. On top of this, 5G is unleashing a much more expansive opportunity for the growth of fixed wireless access (FWA). This growth is geared toward serving high-speed internet over 5G wireless links to homes in dense urban environments, alongside mMTC designed to serve tens to potentially thousands of IoT devices in extremely dense environments. This, of course, is also just the beginning of this shift, as more networks are expanding 5G FR2 functionality with extreme levels of bandwidth and the near-term future adoption of 5G Advanced.

These factors result in changes to where and how data needs to flow through networks that support 5G and the contemporary growth of edge computing services that pair with the development of IoT services. Some of these changes will ensure low latency and the ability to handle the massive numbers of distinct users and variations of traffic supported by 5G. Such changes are already well handled by the 5G standards but will lean heavily on network virtualization features from supporting networks.

However, even the latest networking hardware and algorithms won’t be able to combat the physics of relaying signals from thousands of miles away, resulting in unacceptable latencies for some 5G applications. Hence, there is a growing trend toward building out data centers and edge computing facilities within dense urban areas, campuses, and industrial complexes where the demand for these services is blossoming. Building smaller, more efficient data centers is not an unknown practice, and modern cooling technologies with enhanced efficiencies present a partial solution to the ever-rising costs of electrical supply. One of the more significant challenges emerging is securing real estate within acceptable margins of ROI for various services in the diverse and highly constrained urban markets around the globe. The other challenge is the growth of private wireless networks, often within government and industrial facilities, that will need local data center and edge computing installations, but often with widely diverse requirements.

Accommodating the 5G future, rife with massively distributed IoT devices and the need for edge computing and storage services near users to minimize latency, presents a very different set of challenges than earlier generations of data center requirements. This new set of challenges and requirements will likely require a significant pivot from data center developers and hardware providers to better accommodate the new constraints of more costly real estate and the need to service extreme levels of bandwidth alongside URLLC services as they emerge. Many developers seem to be playing the wait-and-see game. They are merely upgrading their existing hardware and services to better meet more traditional networking demands. At the same time, there is likely room for enterprising developers to take the risk and begin building out micro data center facilities in dense urban environments.