01 March 2004
Web services' new world
Next generation will offer distributed real-time automation, control.
By Jay S. Bayne
The next generation Internet and its supporting Web services will support new classes of applications. One of these classes provides distributed real-time automation and control services. For this class we are interested in the problems of orchestration (coordination and regulation) of performance in a federated enterprise—teams of (semi-) autonomous value-producing entities bound together for higher purpose.
Members of a federated enterprise must (1) be viable and identifiable members of a community, (2) be governed by collectively (and democratically) determined laws, and (3) contribute their individual efforts to coherent ensemble behaviors that characterize the outcomes of the federation as a whole.
As such, there is a higher order logic that defines a federation's mission. As defined in modern democracies, federation objectives include securing the freedom of individuals (enterprises) to pursue their individual objectives while establishing and securing the common good. Federation rules enumerate the powers granted to the federation by its members for their mutual well being.
The three principles enumerated above have direct implications on the architecture and design of distributed real-time automation (i.e., command, or decision, and control applications) applied to grid-connected enterprise.
Consider a multilevel enterprise command and control (C2) hierarchy. The structure depicts the accountability hierarchy comprising two divisions (business areas) and their governing corporate control level. The divisions are each organized into business units, plants, production areas, and production units. At the corporate level the structure includes operations and development functions.
Automation and control services exist at each level. For example, at the unit operations level they provide sequence and modulating controls for motor, power, steam, start-up and shutdown, safety, and production applications. At the intermediate levels control services perform planning, scheduling, maintenance management, and supervisory activities.
At the business unit and divisional levels we find strategic planning, order processing, fulfillment, financial controls, and transportation logistics. As corporations, seeking growth in revenues and market presence, continue to expand through mergers, acquisitions, and alliances, the problems of coordination and control compound.
In the twenty-one years following the January 1983 deployment of TCP/IP within the Advanced Research Projects Agency Network, scientific and commercial enterprises have relied on the Internet to provide transport services for transactional data management applications.
As a result, e-business has blossomed into broad services providing supply chain, electronic banking, information management (e.g., search engines) and "e-tailing" applications. With the advent of the Global Information Grid (GIG) and its emphasis on distributed data and supercomputer support for e-science applications, new Web services are emerging that provide computational resource scheduling, security, and batch control for large-scale collaborative computing activities.
In support of Internet-hosted computing processes, an entire industry has emerged to take advantage of scale economics and provide consolidated and shared communications and computing (ISP) and applications (ASP) infrastructure. This has led to the need to remotely manage network-connected data centers through network operating centers that are operationally similar to telecommunications and broadcast television network operations centers. These centers focus on low-level C2 applications related to infrastructure support.
With the advent of standards and technologies sufficient to host higher level distributed applications, the electronic frontier is ready to address the challenges of defining protocols and services to govern extended enterprises—virtual organizations whose "rules of engagement" are fashioned on, and defined within, the intellectual and commercial (i.e., legal, contractual, management) commons offered by services of the next generation Internet and its enhanced Web. The foundation of such a commons must certainly include a shared concept of value—a means of establishing relative worth between parties engaged in trading information, goods, and services.
Find a value
Enterprises, whether public or private, for profit or not-for-profit, military or commercial, exist to promote and sustain their unique value propositions within evolving financial (equity) and product market conditions. Value production is therefore a continuous and typically time-critical computation of those value propositions. In federations of real-time enterprise, as well as within the legal boundaries of its members, value is often difficult to define and measure. It is difficult within enterprises, because value propositions (and associated value creation processes) are typically spread across organizational boundaries and involve weakly defined business processes. It is difficult across federated enterprise boundaries, because joint (mutual) value propositions must come in terms that transcend participants' unique and often conflicting definitions of asset and supply chain objectives and resources. Hence, any effort to define policies and mechanisms for the management and control of virtual organizations requires the development of a common lexicon, and an agreed upon set of interface specifications to a core set of joint value management services.
The basis of the C2 model relies on the notion of "rational agents," semiautonomous computations that are self-serving (goal seeking) but that participate in alliances as a means of remaining viable, of proactively sustaining their existence. In abstract terms, a viable enterprise is one whose value production units (VPUs), while executing their respective value production processes, are able to cooperate through well-defined interfaces. As such, VPUs define distributed virtual machines.
The essential character of value creation may be expressed in traditional cost-benefit terms, as in income statements and balance sheets. In simplified terms, a VPU is a computation that simultaneously satisfies the requirement to produce positive returns on invested capital assets in a vertical asset chain, while profitably providing goods or services in a horizontal supply chain.
Value production, viewed as a black box, is a process with two clients—investors supplying assets, and customers (clients) supplying demands for goods and services. A given enterprise may include one or more VPUs. A federation comprises two or more enterprises, each containing one or more VPUs.
At a more detailed level, the two-axis model of a VPU supports its asset and supply chains through eight communications ports. Investors provide capital assets at port ai (assets in) that subsequently yield investment returns on port ro (returns out). Customers provide demands for goods or services on port di (demand in) that are subsequently fulfilled on port so (supply out). A VPU supports two subsidiary channels, one for subordinate VPUs and one for supplier (server) VPUs.
Subordinate VPUs are allocated investment assets on port ao (assets out) that generate returns on port ri (returns in). Supplier VPUs are issued demands on port do (demand out) and return their corresponding production on port si (supply in). The horizontal flows may be expressed in terms of cost per unit ordered and price delivered. Vertical flows may be expressed in terms of cost per asset deployed and profit returned. Costs are typically measured in net present value dollars. This model allows VPUs to participate in the production web of a federated enterprise. Each VPU is uniquely identified (i.e., named) by its indexed location vertically and horizontally in the web. Thus, VPU[k,level] is subordinate in the investment chain to VPU[k,level+1], and is a supplier to VPU[k+1,level] in the supply chain.
J. Bayne, Copyright 2003
Enterprise of federations
An enterprise comprising a federation of six VPUs whose aggregation serves the equity and product markets can appear as follows: business unit VPU[2,3] operates three production plants VPU[1,2], VPU[2,2], and VPU[3,2] that together act as a horizontal supply chain to product market customers. Vertically, production areas VPU[1,1] and VPU[2,1] are subordinate to their respective plants, with the entire enterprise collectively serving the equity market through VPU[2,3]. An important question is one of governance—what is an efficient set of management controls to allow this enterprise to effectively and simultaneously serve these two evolving markets?
Enterprises today measure value production through interpretation of operating data produced via transactions in various financial management application subsystems (e.g., ERP) and stored in various databases. There are five major applications found in today's corporations. In this example, the six rings represent the level ("L") dimension in VPU[k,level], where level=0 represents production devices (people and machines), level=1 production units, level=2 production areas, level=3 business units, level=4 divisions or business areas, and level=5 corporations. The radials are representative of primary domains of management C2 focus. The four quadrants represent the major application domains of production, infrastructure, conformance, and asset management. And finally, there are regions that represent application suites provided by vendors of Web-based enterprise software subsystems that are subject to adaptation to GIG standards.
Internet standards and Web services entered this worldview at rings 4 and 5, being initially external to core operations. In addition to inter-enterprise applications, the development of the grid is encouraging its intra-enterprise use within the inner automation rings where proprietary or more traditional communications and computing systems reside.
Within this intrasanctum are traditional distributed control systems for manufacturing; building automation systems for climate, safety, and security; document management systems; maintenance management systems; laboratory management systems; and myriad incarnations of financial management systems, to name a few. With respect to implementing VPUs, we wish to view these evolving applications as formal components in higher level real-time command and control of inter- and intra-enterprise value production processes. For this we rely on the semantics of cybernetics, the science of communication and control in adaptive systems.
J. Bayne, Copyright 2003
A model for intelligent objects shows they are capable of providing reactive and proactive command and control services at all levels of an enterprise.
Such adaptive controllers have been studied and implemented in robotics and discrete, batch, and continuous control systems for years. In this context, the process to be controlled is one or more VPU. The lower loop of the adaptive controller represents the reactive (autonomic) management function, capable of sensing the state of a process as measured through one or more sensors. This input goes to an internally maintained VPU process model through sensory signal processing or sensory perception logic. The process model qualifies the signal processing activity and accepts updates to the database(s) representing the current state and history of the process. Behavior generation logic produces appropriate next state outputs that the object's final controls, or actuators, deliver to the VPU.
The upper loop in this model represents the object's internal proactive management function, supporting volition and the adaptability and intelligence exhibited by the VPU. In addition to having access to sensory input, this value judgment function is able to communicate with other federated controls to declare its intentions and capabilities. It provides methods capable of making qualitative decisions and appropriately biasing the activities of both sensory perception and behavior generation. Taken together, these two interconnected and interoperable loops provide for specific degrees of intelligent automation.
Intelligent enterprise automation, especially adaptive C2, requires specification of and adherence to time constraints.
J. Bayne, Copyright 2003
As such, C2 regimes are inherently real time, within a given enterprise controller as well as among cooperative controls spread horizontally and vertically within a federation. Timeliness and its correlated assumptions about predictability of performance underlie a system's ability to resolve current states and move to new states in time frames required to maintain homeostasis—dynamic stability within changing contexts. As a consequence, imbuing the Internet and its Web services with real-time semantics is at the core of the technical challenge.
This view of federated enterprise, and more importantly the "intellectual commons" upon which freely associative, high integrity, and open commerce may be supported, requires the resolution of a number of legal, commercial, and technicalissues. These issues speak directly to requirements for next generation Internet and Web services. Several infrastructural issues are under review in community standards activities such as World Wide Web Consortium (W3C), the Global Grid Forum, the Internet Engineering Task Force (IETF), and the Object Management Group (OMG). Of the activities in these working groups, the key elements leading to the success of distributed real-time control of federated enterprise include the following topics:
- distributed real-time programming models
- distributed resource scheduling
- distributed system security through authentication, access control, and accounting
Important open issues remain in:
- standardized enterprise operating performance metrics
- standardized enterprise binding services defining and monitoring "rules of engagement"
- standardized enterprise alarm and event services for real-time control
- standardized enterprise operating services for collaborative (coalition) C2 applications