Above the Clouds
What Modern IT Portends
Filippo Balestrieri and Bernardo A. Huberman, Hewlett Packard Labs
The explosion in Cloud computing threatens to make traditional business models and practices in information technology obsolete.
This is because the processing of large datasets using proprietary hardware and software is giving way to commodity software running on computing power that is getting faster and cheaper as time goes on. What once was only available to large sophisticated businesses is now accessible to many with a connection to a cloud service and access to open source software.
These shifts imply that value cannot only come from building and selling large complex systems and computing resources. Increasingly, value is being generated from novel mechanisms and business models that are built around this rapidly commoditizing information technology rather than the machines themselves. Chiefly among them, we envision the rise of outcome-based services, i.e. services that can be purchased at different price points depending on the output quality- level selected.
Currently, the IT market is mostly based on resource-based mechanisms, whereby customers either buy or rent products in the form of servers and virtual machines or, when outsourcing IT processes, purchase services charged by the hour. In all these cases, the trading mechanisms in place do not entail a direct link between the price paid and the benefit earned. Equally important, the risk of experiencing poor results because of wrong choices of IT resources lies completely on the customer side. This is reflected in the fact that current Quality of Service (QoS) contracts are still resource-centric, i.e. they reflect the reliability level in the provision of inputs, rather than being defined in terms of the quality of the output or service performance.
The introduction of outcome-based services will allow customers to interact with technology in a simpler way while giving service providers the opportunity to expand the marketplace and price discriminate among different market segments. This shift towards output-based contracts also means that users will no longer be required to know the technicalities of how IT inputs map into their desired outputs. Rather, they will easily access IT resources while focusing on their core business competences.
The transition from a resource-based transaction market to one in which both resources- and outcome-based mechanisms coexist is not an exclusive peculiarity of the IT market. This evolution has taken place in many other service-providing sectors of the market, ranging from food provision to laundry services.
Three stages seem to characterize that transition. In the first one, the market is exclusively based on the trading of tools; in the second stage, users with better skills in using tools sell their services, and in in the last stage services become standardized and offered at convenient price points with full controls and guarantees over quality levels.
An example of a space in which outcomes-based services have imposed themselves as the norm is the delivery service market. In the past, postal services presented the customer with different shipping options: ground or air. In neither case however, guarantees were provided regarding the effective delivery time. Packages could unexpectedly take long times to reach their destinations and the carrier was not subject to any liability. Recently, delivery companies (think of Fedex or Amazon) started to offer their customers menus with different delivery times at different prices. Outcomes are clearly defined in terms of the waiting time between the moment of purchase and the delivery of a product. Users are not inherently interested in knowing exactly how a package is delivered (which vehicle is used, which route is selected), but rather care about when it will be delivered.
Given the evolving nature of IT, it is interesting to consider how the path towards outcomes-based mechanism is taking shape in the IT market and how outcomes may be defined in the specific context of IT applications.
The technical challenge posed by all outcome-based services is the high level of control that is required over the system performance. Obviously, a service can be
sold through an outcome-based mechanism only if its quality can be clearly identified and measured. Given that, there seem to be still three necessary conditions for outcome-based IT services to become usable. First, access to data on service use under different conditions. Second, control over the quality of the service delivered, so that one can predict the quality level of a specific service as a function of the different amounts and types of resources allocated to it. And third, with this information in hand, an allocation mechanism needs to determine which service quality levels should be offered in order to maximize the effectiveness and utility of the overall system. Once users make their selections, the mechanism has to execute the services requested in compliance with the chosen quality levels. Clearly, these three conditions are strictly connected. For example, a better control over the service quality can be obtained when the data quality is such that predictions are reliable.
The transfer of an increasing number of business and consumer applications to the Cloud implies that Cloud providers have now access to a trove of usage data about different kinds of applications. These datasets of unprecedented size represent unique test-beds for novel machine-learning and optimization algorithms. Advances in engineering, economics and operations research in the analysis of data are allowing a better prediction of service quality levels and the design of more reliable implementation mechanisms (i.e. taking care of what we identified as the second and third condition to enable the sale of outcome-based services).
We believe that we will soon observe several IT applications sold in terms of service completion times instead of by the number of nodes reserved. The development of novel algorithms that combine the pricing and scheduling of jobs and applications is already an active area of research.
Similarly, the data that is collected by sensors and mobile smart-devices may soon be used to launch outcome-based services in new markets. Actually, the whole IoT revolution may find its raison d’etre in making service outcomes measurable and contractible so that companies will start selling outputs instead of inputs. In the future, instead of hiring a taxi and pay the fare at destination, people may be able to purchase directly arrival times; instead of paying for air conditioning and heating system by their electricity consumption, people may be able to purchase temperature intervals on different times of the week.
Outcome-based mechanisms change the language that we use to interact with technology in order to solve specific problems. Indeed, they reformulate business propositions in terms of measurable output quality characteristics. In some cases (e.g. completion or arrival times) such changes are very natural as people are already familiar with what is measured and can immediately associate utility levels with such output quality levels. In other cases (e.g. health services), what to measure is not yet determined but will eventually emerge from the specific needs of users.
It is an established fact that cloud computing has brought a very large dislocation to the IT industry, with new players becoming household names in the market and old ones fighting for their survival. But there is a silver lining to this disruption, for the very technological advances that facilitated these changes are also allowing for a shift in the way IT is sold and consumed.
It is safe to speculate that as outcomes become the coin of the realm in IT a very large and untapped source of users will materialize, if only because what was once perceived as demanding a specialized and arcane language will fade into the background, to be replaced with a business language not very different from the one used in more familiar settings.