Issue No.02 - March-April (2013 vol.15)
Published by the IEEE Computer Society
George Hurlburt , Change Index
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/MITP.2013.24
Although clouds don't freely interact with fellow or competitor clouds today, this is likely to change, leading to larger networks of collaborative, autonomously operating clouds, each optimized for specific services. As this happens, we'll need a framework for understanding this dynamic, nonlinear space representing various networks. Graph theory and the concepts underlying cyberphysical systems can help.
Clouds offer a novel way to revitalize the seemingly archaic concept of time-sharing. Rather than relying on fixed mainframes, clouds thrive in the Internet. Clouds offer impressive efficiencies, unprecedented collaboration opportunities, and economies of scale for all manner of networked users. Yet cloud server farms have enormously costly power consumption footprints and require massive data pipelines to transport transactions and data to and from the servers.
Such connectivity means that clouds rarely stand alone. They currently rely on other equally relevant leading-edge technologies—the Internet of Things (IOT), mobile apps, and big data. Relying on the Internet, clouds both influence how the Internet is used and are influenced by the Internet's use. When connected to banks of remote autonomous sensors, clouds become cyberphysical clouds—that is, specific cloud-based instances of cyberphysical systems (CPSs). Thus, such clouds are citizens of the IOT. 1
Clouds can also offer personal value through mobile apps—for example, reinforcing the idea of Bring Your Own Device. Such apps will continue growing in number and perceived use so long as they can rely on clouds to share data across devices and connect with users. For example, we prepared this article using Dropbox, calendar, and reminder apps working across three different devices. Interestingly, the majority of the cloud-borne transactions for transferring the data to multiple devices were automatic after initial data entry.
Furthermore, clouds harbor big data, be it in small quasirelated datasets or massively large data collections. Other clouds, often relying on Apache's Hadoop open source framework for distributed data-intensive applications, broker big-data transactions and expedite novel views drawing from seemingly disparate datasets.
Although few clouds freely interact with fellow or competitor clouds today, this is likely to change, 2 leading to larger networks of collaborative, autonomously operating clouds, each optimized for specific services. As clouds continue to expand and eventually aggregate, a framework must emerge for understanding the resulting dynamic, nonlinear space embodied in various cloud-related networks.
Graph Theory Can Help
Given that clouds are interdependent with other emergent technologies, the question becomes one of a common denominator against which we can evaluate clouds, the IOT, mobile apps, and big data. The one common element underlying each of these phenomena is networks.
Here, network refers to the interaction of nodes as related by directional or undirected edges. Both nodes and edges can also possess properties, thus adding to the potential of enacting robust discovery mechanisms. Using a Facebook analogy, the term graph extends far beyond the x- y algebraic plots we learn in high school.
We can think of a Facebook profile as properties of a user (node) that's related directly to other users via friendships (edges) and indirectly via properties. The specific links between friends, friends of friends, and so on represent subsets of the larger Facebook graph. For those familiar with graph theory, it's not surprising that Facebook's newest pillar—joining the timeline and newsfeed—is a search tool, appropriately named "Graph." Likewise, graph theory constructs play significant roles in Google and Amazon strategies. Neither could rapidly respond to millions of simultaneous disparate users, each with instantaneously differing interests, without relying on graph-related mathematics. 3
Graph databases traverse edges to discover related nodes. They perform thousands of times faster than relational database systems (RDBSs). These NoSQL tools avoid the dreaded multitude of outside joins necessary to accomplish the same feat in traditional RDBMS environments. Moreover, the graph schema can be recreated on the fly when data is added or deleted. This dynamic frees the developer from the constraints of the rigid relational schemas that doggedly resist change once established. (Neo4j is an open source, multiplatform leader in the graph database arena. This database interacts freely with Gephi, a French open source, multiplatform graph visualization tool.)
Graph theory, with roots in 16th century Europe, 4 is becoming increasingly important. It represents the new mathematics of networking and, as such, yields myriad useful metrics and constructs for rigorously describing networks and the nature of their internal relationships. Given this ability, graph theory applies to almost any digitally networked phenomena. This implies a means of describing and defining all manners of graphs, including those surrounding clouds, IOT instantiations, mobile apps interacting across smart mobile devices and, of course, big data, particularly when unstructured.
Graph theory also serves to describe many other phenomena, including natural networks observed in the hard sciences, such as physics and biology. Significantly, graph theory applies equally well in the soft sciences, including sociology and economics. The significance of these relationships underpin social networks and gives rise to the various direct advertising schemes necessary to monetize these networks. 5
The Next Step: Cyberphysical Systems
If graph theory can be considered a first-order language of network science, perhaps CPSs are analogous to the higher-order constructs surrounding the growing field of network science. The field of CPS focuses on the link between a system's computational and physical elements, including humans. CPS studies strive to integrate concepts involving applied controls, concurrency, communications, interoperability, scalability, complexity management, wireless sensing, and wireless actuation. It also addresses the related fields of reliability, performance, verification, validation, and cybersecurity. 6
As such, the field of CPS brings the concept of "what is a system" full circle, including not only computational software and hardware but also their interactions with their overarching physical environment, well beyond the interaction of software and its computational hardware. These environments encompass numerous fields of endeavor. CPSs clearly extend to robotics and autonomous vehicles with and without humans in the loop. In its call for proposals, the National Science Foundation states: "The December 2010 report of the President's Council of Advisors on Science and Technology 7 calls for continued investment in CPS research because of its scientific and technological importance as well as its potential impact on grand challenges in a number of sectors critical to US security and competitiveness." 8
Taken together, graph theory and CPS research appear to provide the basis for a robust unified framework for understanding both the micro- and macro-level integration of clouds, the IOT, mobile apps, and big data. Applied graph theory and CPS concepts transcend networks of all kinds and simultaneously relate well to larger fields of study touched by computational resources. The framework calls for both unparalleled multidisciplined collaboration and an understanding of applied graph mathematics across related networked phenomena.
Furthermore, it creates a new view of systems integration and interoperability that can embrace existing standards while giving rise to new standards that address practical problem solving in the dynamic, nonlinear space that represents networks of many forms. CPSs, combined with graph theory, could be the "silver lining" for clouds going forward.
George F. Hurlburt, CEO of Change Index, has an abiding interest in applied complexity analysis. Contact him at firstname.lastname@example.org.
Jeffrey Voas is an associate EIC for IEEE IT Professional, is on the editorial advisory board of IEEE Spectrum, and is on the editorial board for Computer. Voas is an IEEE Fellow and a Fellow of the American Association for the Advancement of Science (AAAS). Contact him at email@example.com.