SAVI will pursue five research themes:
- Smart Applications. Investigate and design reusable frameworks for rapid deployment of high-potential applications enabled by virtualized application platforms.
- Extended Cloud Computing. Investigate and design adaptive resource management to provide effective, efficient and reliable support for applications across all elements of the application platform while attaining economic, sustainability, and other high-level objectives.
- Smart Converged Edge. Investigate and design smart-edge converged infrastructure that uses virtualization and cloud computing principles to dynamically support multiple network protocols and high-bandwidth, latency-sensitive applications.
- Integrated Wireless Optical Access. Investigate and design high-bandwidth, energy-proportional adaptive virtual access networks based on dense small cells and dynamic optical backhaul.
- SAVI Application-Platform Testbed. Design and deploy an application-platform testbed that includes integrated wireless/optical access, smart converged edge, wide-area network connectivity and datacenters; Use testbed to test future Internet architecture alternatives and smart applications; Develop and provide tools for an ecosystem of open source projects.
SAVI Research Plan Structure
Consider the following scenario: The Blue Jays are in the final game of the 2020 World Series and they are mounting a comeback in the last inning. Reporters and pundits report on the game through live audio and multi-angle video streaming feeds that are followed by the public on their 10G iPads and Greenberries. Fans in the stadium and throughout the city share their reactions and commentary using social networking apps and contribute their own audio and video streams from personal devices to shared media repositories. Using speech and video recognition, audio and video streams are tagged in real time with timestamps, location, and by subject. Viewers anywhere create one or more customized views, e.g., from home plate, of the MVP, etc., using the latest release of the Kaleidoscope app.
As the Jays win the game, data traffic peaks. The “smart edge” activates the maximum number of pico-cells to provide capacity to carry and backhaul massive data streaming content wirelessly. Local virtual circuits route data to storage in the edge and beyond. The smart edge must ensure that local security personnel continue to have their own secure and dedicated channels for the purpose of law enforcement. As the crowd moves to local restaurants and bars to celebrate, monitoring in the smart edge and core platforms detect and predict decreasing volumes of traffic, and the power of pico-cells is dynamically and progressively turned off. Stadium traffic is reduced to fit in one wavelength and power is turned down on lasers and receivers on the local backhaul. In contrast, dormant pico-cells in restaurants and bars in the area are powered up and backhaul capacity to core datacenters is increased to accommodate the demand as fans turn to Kaleidoscope on giant screens to relive their favourite moments.
The above scenario illustrates the agility and flexibility required in future application platforms. The SAVI Network will address the design of future application platforms built on flexible, versatile and evolvable infrastructure that can readily deploy, maintain, and retire the large-scale, possibly short-lived, distributed applications that will be typical in the future applications marketplace. The figure above shows the infrastructure elements of an application platform. Future users will typically access the application platform through a mobile device that connects to a ubiquitous very-high-bandwidth, integrated wireless/optical access network. The application platform provides connectivity to services that support the application of interest. Many services will be supported by massive-scale distant datacenters located at sites of renewable energy. Other services will require low latency ( alarms in smart grids, safety applications in transportation, monitoring in remote health) or processing of large volumes of local data (e.g., video capture in lecture rooms) provided by converged network and computing resources at the “smart edge” of the network, such as the premises of service providers.
SAVI considers the computing cloud to include the smart edge, and it will investigate the interplay between the smart edge and remote datacenters in the delivery of applications. SAVI will investigate the hypothesis that all computing and networking resources, including the integrated wireless/optical access, can be virtualized and managed using infrastructure- and platform-as-a-service principles. A key objective will be to develop application-development methodologies and the intelligent and automated management and control systems that will transform the application platform into a powerful and highly flexible enabler of a broad range of applications.
The SAVI research plan involves five themes addressed by a team of researchers in nine universities.
1. Smart Applications. The theme will develop reusable frameworks for three classes of applications that high potential impact in future application platforms: 1) Data-intensive applications, whose design requirements include high performance on the run-time interaction of users accessing data; 2) Mobile applications, architected in an event-driven style, with support for a multitude of “typical” event types, including frequent location updating, P2P communications and services, and on-the-cloud-workflow management; and 3) Media applications, architected in a highly distributed style to support flexible, redundant caching, with transparent and efficient transcoding and streaming services. The theme will develop an autonomic-management framework that will guide resource provisioning during application deployment, monitor application behavior, and re-provision application resources to ensure cost-efficient high-quality performance for the application and high utilization for the platform.
2. Extended Cloud Computing. This theme will investigate mechanisms to develop adaptive software systems that change their behaviour and structure to cope with changes in environment and user requirements. Our focus is on adaptive resource management: dynamic allocation of software, hardware, and network resources based on runtime models. A dynamic resource management system will be developed to provide efficient resource utilization, low-power consumption and improved ability of the system to adapt to changes in workloads. With resource virtualization, distributed applications are deployed on “virtual networks” of converged computing and communications resources that span datacenters and the smart edge. In particular, we will investigate the design of storage and data services for adaptive, multi-tiered applications such as SAVI’s. Automated resource management is achieved by adjusting the capacity allocated virtual networks according to predicted workload and observed performance. The automated resource management systems in all SAVI themes require a framework for characterizing and forecasting application loads and for monitoring QoE. SAVI will develop methods to measure and characterize the service flows generated by classes of applications.
3. Smart Converged Edge. This theme will develop a programmable converged network and computing infrastructure for the smart edge that provides responsive and high-capacity virtualized resources and services close to the user. The smart edge will provide virtual computing, networking, programmable hardware processing, and storage that complement the resources provided in remote datacenters in support of applications. The implementation of routers using virtual resources will be investigated with a focus on energy efficiency, scalability and programmability. We will investigate how the virtualized infrastructure at the edge will support: proposed Future Internet protocols; energy-efficient networking; novel wireless processing and access protocols; real-time media processing; intelligence through analytics; content caching, content routing, publish-subscribe, and other multi-layer protocols and services. Algorithms will be developed for the creation and management of reliable and efficient virtual networks in the smart edge, as well as for the scaling of resource allocations as demand varies. The theme will develop online methods to relate QoE to key performance indicators, and to adaptively allocate resources to achieve target values of QoE.
4. Integrated Wireless Optical Access. To support much more demanding content-oriented or latency-sensitive applications, the access infrastructure in future application platforms must provide ubiquitous, very high bandwidth wireless access. The theme will develop virtualized radio-over-fiber technologies that can be integrated with virtualized and programmable, very high bit-rate, dense small-cell access networks. The theme will design adaptive runtime resource management methods that deliver target QoE/QoS levels to applications while adjusting the spatially-available capacity to the current demand in an energy-proportional manner. Novel software-defined radio (wireless signal processing, networking protocols), content caching and high-bandwidth and low-latency services will be explored, leveraging the smart edge in Theme 3. Services to support sensor and environmental networks will be considered.
5. SAVI Application-Platform Testbed. A major obstacle to innovation in network aspects of application platforms is the inability to test and deploy new Internet protocols at scale. The SAVI research plan includes a shared activity to build a testbed that enables experimentation in future application platforms and future Internet alternatives. The theme will develop a smart-edge cluster that provides virtual resources in support of networking and applications. The SAVI control and management software developed in collaboration with Themes 2 and 3 will be used to create virtual networks. Researchers will obtain virtual nodes at different SAVI smart-edge clusters and use R&E connectivity to build networks that support experimentation at scale. The SAVI clusters will incorporate open source software and hardware (OpenFlow, NetFPGAs) from other testbed activities, and they will include federation methods to interconnect other testbeds. To enable experiments that require high performance, the theme will develop tools to facilitate the programming of FPGA resources. Very high-bandwidth dense small cells from Theme 4 will be built and integrated with radio-over-fiber systems, and used to test the dynamic allocation of energy and capacity to small cells as demand varies. Theme 1 will provide APIs for the SAVI application platform and tools for the creation of targeted application classes by students and members of the academic community.