Bridges vol. 41, October 2014 / OpEds & Commentaries
By Schahram Dustdar, Michael Vögler, Sanjn Sehic, Soheil Qanbari, Stefan Nastic, Hong-Linh Truong
Vienna University of Technology
Smart cities encompass services in diverse business and technological domains. Presently, most of these services are delivered through domain-specific, tightly coupled systems provided by specific vendors, which entail limited scalability and extensibility. This joint research lab proposes an Internet-scale service delivery that addresses these limitations and encourages the creation of novel services based on a domain-independent, cloud-based service-delivery platform.
In a joint research lab, the Distributed Systems Group (DSG) of the Technical University (TU) of Vienna and Pacific Controls are working together to push the frontiers of cloud computing, pervasive computing, data management, and the Internet of Things, as shown in the figure.
By creating social and physical environments conducive to innovation and knowledge transfer, researchers and practitioners are working in partnership to create novel service platforms and technologies and continuously transfer them into business values. The purpose of this collaboration is to enable the development and evolution of smart cities through software and service techniques. It aims particularly at providing strong sustainability support over the whole life cycle of the various complex systems in smart-city environments.
Current smart-city services are typically provided in single domains – for example, building management, transportation, and health care, among others. With such services, domain-specific application requirements drive all system-component design and determine most technical choices, ranging from sensors and smart devices to middleware components and computing infrastructure. The service-delivery process is rigidly orchestrated by domain solution providers who develop applications and integrate subsystems from various vendors. This model leads to many closed vertical systems with tightly coupled hardware, networks, middleware, and application logics. Scalability and extensibility are intrinsically limited in such systems, and the closed relationships between stakeholders stifle the creation of novel services.
Service delivery for smart cities requires a new methodology aimed at delivering open and scalable smart-city services by encouraging collaboration between stakeholders in the Internet of Things and clouds. It exploits the prevalence of computing resources and software services distributed on the Web to break up the closed vertical service-delivery model.
Smart-City Services – Two examples
First we will analyze two representative smart-city services and examine the limitations of the closed service-delivery model.
Building Energy Management Systems
Managing energy for buildings is one of the most important services in smart cities. A building energy management systems (BEMS) project typically starts when a solution provider surveys the target building, which can already exist or still be in design. A project’s scale can range from a small building to a large campus comprising various building types. So the number of devices, the volume of data that will need processing, and application complexity can vary significantly. After surveying and designing for the specific building, the solution provider will acquire suitable hardware devices from original equipment manufacturers (OEMs), integrate them into an infrastructure solution, and develop analytical and control applications.
This process produces many vertically isolated BEMS (often referred to as “silos”), which leads to two problems. The first is maintainability: The more silos that are provisioned, the more system instances the solution provider must maintain. System maintenance becomes a particularly painful process – hardware devices are monitored in separate systems, and software instances must be updated separately and tested on site with specific hardware configurations. The second issue is extensibility: Many campuses and building compounds expand continuously to accommodate new users; deployed BEMS might need to manage more buildings and facilities. In any case, BEMS ought to scale up accordingly.
In the current silo-based service-delivery model, such expansion could require the solution provider to reconfigure the whole system from the bottom up or to add new isolated BEMS because of tight coupling among devices, middleware, and applications. Furthermore, the energy consumption data that individual BEMS collect are largely underutilized because data storage and processing modules are isolated. Today, a painstaking data cleaning and integration process is a prerequisite for conducting fine-grained energy consumption analysis on a large scale.
In brief, the BEMS exemplifies a vertical service-delivery model, by which a single solution provider is generally responsible for provisioning and maintaining the entire solution throughout its lifetime. This is the dominant model by which most current smart-city services are delivered. Scalability and extensibility in this model are inherently limited.
Public events constitute an essential part of urban life. Some events reoccur regularly, such as yearly city marathons or national day parades; some might be ad hoc, such as demonstrations. An event at any scale has effects on city dwellers. Participants or visitors want to know how to reach and leave the location. Those who aren’t interested want to know how to avoid the event. Public services should be ready for expected or unexpected situations. Event organizers must address all these concerns.
A typical event-organization service is composed of data collections, organization plans, notification channels, and contingency plans. It runs through four phases in normal situations – planning, preparation, operation, and finishing. In addition, contingency operations might be carried out if unexpected situations arise. The information required for an efficient organization application is diverse; it can include relatively stable information such as maps, public transportation routes, communication channel capacity, and facility accessibility, as well as real-time data such as traffic, weather, public transportation status, and parking lot occupation. Although most information of this type is available through existing public services, these services are isolated and domain-specific. They operate their own information infrastructures, process the data in house, and publish them via various public channels.
The key challenge to developing event-organization services is accommodating the event specifics (scale, local resources, processes, safety concerns, and so on) by properly collecting domain-specific services and information sources. Developing an event-organization service requires significant effort, given that it might be used only once or may need to be updated periodically for regular events. Furthermore, the computing resources required to provide such services are needed only during the events. Thus, the effort required to deliver such a service is justified only if the development and provisioning are efficient and cost-effective.
Public-event organization is a case of third-party service delivery: An application provider acquires access to existing IoT services and other information sources to develop applications for specific purposes. These information sources are highly diverse, ranging from public services such as public transportation status to commercial services such as mobile networks. The provider’s focus is on developing the application logic because it doesn’t own the information sources or, in most cases, doesn't have direct control of them. However, the provider must provision computing resources to ensure quality of service (QoS). Third-party smart-city applications, particularly those needing to incorporate multiple IoT and Web services, are still uncommon due to the challenges exemplified by this case.
Stakeholders in Smart-City Service Delivery
At the center of Internet-scale smart-city service delivery are domain-independent service-delivery platform providers, who present a new type of Platform as a Service (PaaS) offering that integrates IoT devices and infrastructures, processes data from a large amount of distributed data sources in real time, and lets applications employ both IoT and cloud resources on demand. The management of both IoT infrastructure and cloud resources is hidden from application providers. Platform providers must ensure the required provisioning of computing resources involved in service delivery. The platform also provides cloud services, including service metering, billing, and tenant management that will let stakeholders share resources and establish flexible business relationships.
Such a platform’s emergence will directly influence traditional domain-specific solution providers. With a cloud-based platform, solution providers can leverage cloud resources to integrate IoT infrastructure and develop domain-specific applications, thus enabling virtualized vertical solutions, or virtual verticals. In virtual verticals, solution providers can reuse software services on the cloud and scale up services without investing in the computing infrastructure. In addition, other than the traditional role of providing vertically integrated IoT solutions, solution providers can also provide IoT Infrastructure as a Service (IaaS) on the cloud to open IoT device capabilities to third-party application developers.
The platform will also benefit cloud application providers who specialize mainly in Web and cloud application development. The service-delivery platform lets these providers access IoT services to create novel applications for users. Application providers won’t need domain-specific knowledge for managing IoT infrastructures because such infrastructures’ capabilities are provided as services on the cloud, and the platform facilitates the important components for service delivery. Thus application providers can focus on application logics and enjoy on-demand use of both cloud and IoT resources.
Collaboration between the aforementioned core stakeholders will enable open and scalable service-delivery models for smart cities.
The virtual-vertical model corresponds to traditional vertical solution development: Solution providers integrate IoT infrastructure (devices, networks, and so on) and develop application logics. The key difference is that the virtual verticals use computing resources and other necessary software services, such as data service and metering, on cloud platforms.
At first, IoT infrastructures are integrated through virtualization, which is a set of commonly applied techniques for opening service interfaces to access device capabilities. A solution provider registers these IoT services to the platform for further use. Then, the provider can either offer them as IoT Infrastructure as a Service (IaaS) by configuring the services’ metering and billing schemes, or use the services to develop virtual verticals. In the Platform as a Service (PaaS) paradigm, virtually isolated system environments are ensured for applications via tenant management and application environment configuration. In the configured application-hosting environment, applications can scale up and extend their capabilities by employing the resources under an agreed tenant relationship, and the elasticity of cloud resource provisioning can help the virtual verticals deal with fluctuating user demands. Finally, applications can also configure their metering and billing schemes to offer their services on the cloud.
This model is widely applicable to most existing vertical solutions. The first use case domain – BEMS – is a typical example. The platform can give each building a virtually isolated operation environment that is based on the IoT infrastructure deployed in the building. The energy-management system is delivered on the cloud and has access to buildings under its management, while sharing computing resources and data processing services with other services. This also allows aggregation of data for analysis from multiple management domains.
In traditional, vertical IoT service delivery, third-party application development isn’t common due to tightly coupled system architectures and inflexible resource provisioning. On the smart-city service-delivery platform, third-party application developers can use platform services, including data and virtualized IoT services, thus mitigating the difficulties that result from using IoT capabilities in third-party applications. The usual concerns of device maintenance and data acquisition in IoT services are uncoupled from application development, so that application providers can focus on their business logics and quickly deliver new services.
Application development starts with discovering available IoT services and acquiring access, which might require application and IoT service providers to reach an understanding on service-level agreement (SLA) and billing terms. Similar to the virtual-vertical model, applications can have their own hosting environments configured with necessary virtualized IoT services, platform services, and computing resources. Furthermore, application providers can configure flexible usage models and billing schemes for applications so that the cost and revenue associated with them are monitored in real time. In this way, the platform fosters an environment for creating and delivering new applications as services.
Public-event organization is a typical case for using this model. Lower-level devices and domain-specific capabilities are publicly available. Application developers can access public information through the platform and acquire the specific data related to the event. Application providers are relieved from having to provision computing resources and from surging user demands. Furthermore, the PaaS paradigm gives specialized IoT solution providers an environment in which to serve multiple users through virtualization. For example, providers can use real-time public transportation monitoring in other applications.
Another core activity in the domain of Smart Cities is addressed by the interdisciplinary Doctoral College entitled Urban Energy and Mobility Systems (URBEM). Its primary focus is the research and development of scenarios for a sustainable, affordable, and supply-secure city based on the example of Vienna. In cooperation with the Vienna Public Utilities Company, a distributed interactive analytical system is being developed, enabling domain experts in the areas of mobility, energy, economics and buildings to visualize their models in an interactive 4D environment. Such a system not only faces the challenges of integrating and handling diverse, very large, multidimensional data but also coordinating temporal diverse complex models to present a coherent result. The goal is to develop a novel foundation for the development of future-proof energy and mobility systems.
Schahram Dustdar is a full professor of computer science and head of the Distributed Systems Group at the Vienna University of Technology. His research interests include Services/Cloud computing, Internet of Things, complex and adaptive systems, and context-aware computing. Dustdar is a member of the Academia Europaea, and an ACM Distinguished Scientist, an IBM Faculty Award recipient, and editor-in-chief of Computing (Springer). More information at: http://dsg.tuwien.ac.at/Staff/sd