The choice of organizations: up in the cloud or on the edge

In the past fifteen years, cloud computing has come to the fore as the paradigm of choice for managing information systems in businesses. A centralized, shared data center infrastructure makes it possible to achieve significant economies of scale, at the same time lightening the load for individual organizations in investments for maintaining proprietary systems. Since the computing power and speed offered by cloud architectures outclass on premise for most organizations, the cloud is destined to remain a cornerstone of contemporary computing architecture. This also thanks to the proliferation of high-speed connectivity.

And yet in recent years the expansion of the cloud is losing speed in part by a number of technological bottlenecks: specifically high latency caused by huge consumption of band width. The reaction to this is an alternative yet complementary trend toward decentralization. In other words, generating, collecting and processing data directly in close proximity to the things or devices that produce these data. This is what’s known as edge computing, which is growing thanks to 5G networks and the explosion of Internet of Things technologies – “things” embedded with sensors that are capable of detecting and processing data autonomously.

By some estimates, in the next five years the market for edge computing could be worth anywhere from 175 to 215 billion dollars. The advantages for organizations of decentralizing computing systems are myriad: reduced latency and the possibility to make immediate data-driven decisions, lower costs, enhanced resiliency, and a lighter load on network infrastructures. But despite all these benefits, before adopting a strategy “on the edge,” companies must carefully consider the decision, not only with regard to technical and security concerns, but also at an organizational level. In fact, prerequisite to defining an edge strategy is analyzing all the vertical and horizontal relationships that underpin the organization’s value chain.

For fifty years now the debate has raged on about the impact that different computing infrastructures can have on the organizational structures and decision-making processes in companies. In general, studies on this question suggest that computing infrastructures alone, in their various shapes and forms, do not necessarily make decision making more or less centralized. The decisive factor instead is the organizational context.

So in our study, we crossed different types computing infrastructures (centralized and decentralized) with different configurations of company decision-making processes (centralized and decentralized). Our findings reveal four possible organizational architypes, illustrated in the organizational and computing matrix below (Figure 1).

Figure 1: Organizational and computing matrix

The first possible archetype is the so-called canonical organization, which combines both a centralized computing infrastructure with a centralized decision-making system. In this context, all data are channeled into a single core entity where they are processed; this is also where all decisions are made. The people on the edge of this system simply collect data in the initial phase, and then implement the resulting decisions. This kind of archetype – by far the most common - is good fit in many contexts with low uncertainty and high stability, as well as where greater flexibility is needed.

When decisions are made centrally but the computing infrastructure is decentralized and distributed around the edge, we have an autonomous centralized organization. With this archetype, individuals on the edge do not have to make decisions autonomously, but they do have to be capable of exploiting the potential of edge computing. A typical example of this archetype is a predictive manufacturing system, which has machinery embedded with sensors that can detect and anticipate possible anomalies.

The third archetype characterizes organizations that encompass a plurality of entities with decision-making autonomy which are connected to a single, highly-centralized computing infrastructure: the so-called platform organization. This is a particularly appropriate archetype in contexts where activities are performed that are similar in theory, but which require a level of personalization and decision-making autonomy.

Lastly, in a decentralized autonomous organization, both the computing infrastructure and decision-making mechanisms are decentralized. In other words, there is a network of autonomous entities that can function independently while working in a shared network in which information circulates freely. An example of this kind of archetype is Bitcoin, the first blockchain protocol ever developed, based on completely decentralized, distributed peer-to-peer architecture.

Contrary to what we might think, a centralized organization does not necessarily correspond to a centralized computing architecture, or vice versa. The decision to opt for a cloud paradigm or to embrace the emerging trend toward edge computing has to be based on a careful analysis of the compatibility of company processes.

• To fully exploit the potentialities of edge computing, the devices “on the edge” must have the capacity not only to generate data, but also to process them autonomously. If organizations on their own does not have the necessary computing power, they have no choice but to resort to an external centralized provider.
• An edge strategy only makes sense if edge devices are able to transmit data quickly and extract and process all information needed to make relative decisions autonomously.
• To guarantee the security and reliability of the system, clear accountability must be established in case edge devices make the wrong decision. Is the producer responsible? Or the software the device runs on? Or the company’s security architecture? If an organization wants to embrace an edge strategy, these questions must answer be answered ahead of time to avoid having to take sole responsibility for possible failure.