Multimedia content providers have faced an exponential growth in digital content in recent years, and this explosive trajectory is likely to be sustained in the future. Information lifecycle management (ILM) using multiple tiers of infrastructure at varying performance and operating cost levels is a natural solution to this growth and the shifting popularity of digital content over time. However, content migration across tiers has traditionally tended to be expensive because of extensive planning, debugging, and reconfiguring.

Furthermore, firms (both content providers using in house facilities as well as infrastructure-as-a-service— IaaS—providers) therefore rarely tended to archive data and thus overprovisioned the top-of-the-line storage infrastructure (such as the solid state drives or SSDs) to ensure scalability and prevent service disruptions. Putting all of a firm’s multimedia objects in such a highly accessible high-performance storage and delivery system may assure exceptional quality of service for all objects at all times. However, this extreme strategy ignores the temporal and media-specific characteristics of digital content and assumes the same revenue earning potential for all objects across the board, thereby being inefficient from a profit maximization perspective. This practice leads to significant inefficiencies in terms of resource utilization, cost, returns on infrastructure investments, and ultimately net returns of the firm.

Moreover, Du et al. (2014) carried out a longitudinal empirical study to evaluate the proposed migration and revenue-management strategy. They collected data on 1,334 unique videos from Amazon VOD over 13 weeks and used these to construct the complete digital media network for each week and to approximate their expected direct hits. They have developed a joint strategy by which a tiered service provider such as Limelight can set quality of service–differentiated infrastructure prices and a media content provider such as Amazon VOD can allocate digital content across tiers. They propose a bi-level model where both parties maximize profits and develop a polynomially-solvable optimal algorithm. They model two primary effects—direct access differential revenue effect caused by the tiers and traffic-generating effect caused by the media objects— which together lead to a traffic convergence effect on any node, resulting in direct and indirect marginal revenues from a node.

Finally, some key managerial implications emerge from this study. First, sellers of tiered services would do better to maintain a narrow technological gap between the two tiers; because of the highly connected nature of digital media objects, a larger difference in quality of service would mean greater erosion in network wide revenues. Second, if the buyer and seller are interested in forming a vertically integrated alliance, the bi-level solution can serve as the basis for negotiation and the evaluation of the transaction costs in that process. Finally, videos in the long tail can justifiably be placed in Tier I, as they enable discovery of other videos; even though their direct access effect is relatively lower, their traffic-generating effect has larger network wide revenue implications, often more so than more popular items.

Resources:

Du, A.Y Das, S. Ram, D. Gopal, Ramesh, R. (2014). Optimal Management of Digital Content of Tiered Infrastructure Platforms. Information Systems Research 25(4): 730-746.