Internet Economics: Models, Mechanisms and Management

This chapter covers the evolution of the Internet in its historic setting, the emergence of Internet institutions, technologies, architectures and services. From this an Internet industry developed as taking major dominance in the telecommunications world expanding to a global context. Key economic issues involve Internet pricing, Quality of Service (QoS) provisioning, congestion management, Internet regulation and the backbone network technologies such as Asynchronous Transfer Mode (ATM) in view of traffic management and congestion control. The relation between service discipline and the varieties of the bandwidth-buffer tradeoff is also discussed. It shows improvements being made to effectively operate and manage networks that may support performance and guarantees in serving heterogeneous users with a wide range of requirements.

The Internet is framed as a market design for information with major actors supplying and demanding information, and supply and demand tending toward equilibrium, as a paradigm of a distributed computational system. Concepts of information flows and delivery provisions would reflect game-theoretic models of trade and exchange economies where the mechanism (design) of information flows would adopt the behavior and outcome of mathematical queueing systems. Of particular interest are large scale systems of the Internet type that act and perform as decentralized distributed systems of multi-level control. A prototype economic optimization model is put forward for a network economy with variation in traffic classes, utility parameters, queueing types and equilibrium approaches.

We consider specific examples of network economies with network routing and transaction processing. Any of these examples capture a structural model of the network economy with Pareto optimality and price equilibrium for agents competing for resources from emerging suppliers. A routing algorithm establishes the dynamic nature of of session arrival and departure.

We focus on economic management of web services in distributed multimedia systems. We dig deeper into mechanism design approaches tracing them back to classical economic mechanisms of market designs and information economics which I refer to as Hayek-Hurwicz mechanism design – due to Austrian-British economist F.A. Hayek and American economist L. Hurwicz. The basic idea of market agents as computationally efficient human agents induced by incentive compatability and selfishness have been rediscovered and reapplied in rigorous methodological form by computer scientists merging algorithmic game theory, computability and network complexity. Distributed algorithmic mechanism design (DAMD) for internet resource allocation indistributed systems is akin to an equilibrium converging market based economy where selfish agents maximize utility and firms seek to maximize profits and the state keeps an economic order providing basic public goods and public safety.A distributed algorithmic mechanism design thus consists of three components: a feasible strategy space at the network nodes for each agent or autonomous system, an aggregated outcome function computed by the mechanism and a set of multi-agent prescribed strategies induced by the mechanism. A distributed algorithmic mechanism design being computationally efficient in a large decentralized Internet economy is a powerful paradigm to substantiate claims by Hayek (1945) that an industrialized economy based on market principles has an overall better output and growth performance (static and ynamic) than socialist type economies of a similar nature and scale. Best economic coordination through markets producing maximal social welfare is supported by computational efficiency in computer science. Applications relate to a data management economy.

Broadening the criteria for QoS performance and guaranteed service through reputation systems hinge on trust and belief. As the Internet matures in promptness, reliability, accessibility and foremost security for a certain targeted QoS level, this is termed as ‘Generalized QoS level’. GQoS includes emphasis on security, reputation and trust. Invoking economic design principles in the present day Internet have provided measurable ingredients for QoS as they identified performance criteria such as average response time, maximum response time, throughput, application failure probability and packet loss. QoS can be affected by various factors, both quantitative (i.e., latency, CPU performance, storage capacity etc.) and qualitative that proliferate through reputation systems hinging on trust and belief .As the Internet matures in promptness, reliability, availability and foremost security for a certain quality service level targeted, it embraces a GQoS. level.

A new property of Internet enabled communication and computation opens up a new business model of ‘plug and play’ that is intrinsically linked to the interactive social and commercial use of websites through the World Wide Web (WWW), i.e. two-sided or multi-sided platforms. We analyze a model of platform operations that involves a sequential decision process very much alike a dynamic programming algorithm (DPA) for selecting functionalities of the platform. This could serve as a simple approximation procedure for building the optimal design of a platform. The platform business in a vertically integrated supply chain as well as toward product development through platforms rather than pipelines is a good example of facilitating ‘increasing returns mechanisms’ (IRM), as one can follow the evolution of the Amazon platform in a commercial context but also on the growth of social media like Facebook and LinkedIn. Such a business growth would not have been facilitated without a dedicated, universal, easily accessible and low cost network economy provided by the Internet. From an economic and business perspective, the network platform business utilizes direct or indirect network effects to attract customers and facilitate network economies, and determine how to switch from a ‘pipeline model’ (product line model)to a ‘platform model’. Some business and social network platforms (eBay, Facebook) have been able to harness network effects to fuel truly continuous growth. On the other hand, platform businesses, even without fixed costs or economies of scale, need to acquire critical mass through intensity and scale after they are launched to survive and grow. We may observe ‘tipping point’ effects generated through ‘chicken and egg’ problems in platform building processes.

The Internet in the form of expansion of things, or everything, the Internet of Things (IoT) shows how Internet could lead to a radical dimensional and combinatorial type expansion. The Internet of Things (IoT) embraces connectivity up to a scale of aggregate capacity and complexity. It could evolve into a network where not only each node would be a computational device but by itself, it would also be an intelligent computing device that could replace human supervision and control through Artificial Intelligence (AI). The Internet of Things (IoT) as being embedded in the ‘Industrial Internet’ is a new paradigm shift that comprehensively affects computers and networking technology. This is also recently referred to the buzzword ‘Industry 4.0’. This technology is going to increase the utilization followed by Bandwidth of the Internet. More and more (intelligent) devices in this network are connected to the Internet through various combinations of sensor networks and interactive machine learning.

For observing a huge online data generation through an expanded Internet (IoT and the Industrial Internet) one can use the flux of information for monitoring, predicting, controlling and decision making. During the time of applying methods of statistical inference and statistical decisions some 70 years ago, information derived from data collection was considered costly. Models were built where information was linked to payoff relevance of a decision making criterion (utility or payoff function), therefore statistical information was handled to satisfy these criteria. What really is subsumed under ‘Big Data’ (BD) qualifies under a few main characteristics: (i) BD is primarily network generated on a large scale by volume, variety and velocity and comprises large amounts of information at the enterprise or public level, in the categories of terabytes (1012 bytes), petabytes (1015) and beyond of online data. (ii) BD consists of a variety and diversity of data types and formats, many of them are dynamic, unstructured or semi-structured and are hard to handle by conventional statistical methods. (iii) BD is generated by disparate sources as in interactive application through IoT from wireless devices, sensors, streaming communication generated by machine-to-machine interactions. The traditional way of formatting information from transactional systems to make them available for ‘statistical processing’ does not work in a situation where data arrived in huge volumes from diverse sources, and where even the formats could be changed.

From Internet induced economic effects in micro-economic structures, i.e., on enterprise and industry levels, we now address network effects on a macro scale involving productivity, growth, and the business cycle. The network effect is strongly facilitated by computerization and information technologies (ITs). Ubiquitous computerization and digitalization increasingly pervade many sectors of the economy, and communication by network technologies through the Internet (the network is the computer) as a strong catalyst. Eventually, through this synergy, most sectors of the economy will be impacted by network effects. Thus, networking and computerization have far-reaching impacts on the pace and path of the economy but they could also make the economy more vulnerable to economic shocks and security breaches. We address three important issues: networks and productivity, endogeneous growth and increasing returns. The relationship between technology and productivity used for the United States on the economy or sector level, found little evidence of a relationship in the 1980s. Capital IT investment between 1977 and 1989 rose several hundred per cent but was barely reflected in a rise in output per worker. There was this famous saying by Nobel prizewinning economist Robert Solow (1987): “You can see the computer age everywhere except in productivity statistics”. In the 1990s, such a positive relationship on the firm level was established empirically. On a short-term basis: one-year difference in IT investments vs one-year difference in firm productivity should be benchmarked by benefits equal to costs. However, benefits are supposed to rise by a factor of 2–8 in forecasting future benefits through productivity growth.