You are here

Research, Education and Government Use Cases

Introduction  

Research, education, and government networks can benefit from SDN in multiple ways. The problem sets, value propositions, and markets vary slightly. For instance:

  • Universities may need specialized high performance networks
  • Municipal or state governments may be able to improve the quality of life via machine-to-machine applications
  • National Research and Education (NREN) networks may have multiple use cases over large geographical areas

 

An effective solution may involve a pure OpenFlow-based approach or a hybrid approach including existing protocols. While organizations (whether service providers or enterprises) may be augmenting their existing networks with new options, research or other greenfield networks often have comparatively limited nodes and flows but need high performance and exacting control, and may be able to opt for a solution entirely based on OpenFlow. In all cases, an SDN framework can be used to centralize network control and management functions, resulting in enormous gains in service flexibility and reliability.

Campus or smaller networks may lend themselves to a pure greenfield approach as well, particularly for research institutions that need unusually high performance or a network that is highly intelligent and flexible. Examples include university science or computer and information systems (CIS) departments needing the power of big data analytics.

In local governments, there is a movement towards building "smart city" infrastructures where SDN, Cloud and Internet of Things (IoT) are used to make metropolitan areas more livable and sustainable. This type of application may lend itself to a large number of sensors providing real-time data in an IoT application. 

NRENs are effectively service providers dedicated to supporting the needs of the research and education communities within a country, and may in some cases be connected by pan-continental networks. NRENs have WANs that must integrate the packet, optical transport and software layers of the network while leveraging analytics to intelligently automate and orchestrate operations between them.

Like their commercial counterparts, NRENs can benefit from automated service delivery and network resource optimization. They have somewhat more freedom, however, to approach these problem sets from a greenfield perspective.

The Challenges

Campus networks, facing major challenges, are ripe for innovation. These changes require greater agility and performance than legacy solutions can provide. Demanding applications need improved performance, reliability and security while preserving traditional network connections on existing hardware.

A similar problem exists in regional government networks, where disparate hardware technologies and protocols inhibit the goal of seamless connectivity. Managing hundreds or thousands of sensors needed to create a smart city requires central control and convergence of these disparate networks.

Furthermore, a smart city consists of a wide variety of dynamic applications that require access to geographically distributed sources of information. In addition, each service has its own specific needs in terms of bandwidth, availability, latency and other QoS parameters.

Finally, NRENs need improved network monitoring and resiliency, to reduce the operations cost of service management, improve recovery times, and provide more efficient usage of their network capabilities.

Why OpenDaylight?

The OpenDaylight (ODL) platform includes the tools to address the unique problems of research, education and government networks.

ODL enables the innovation needed in research and education, exploiting global network views for traffic engineering and security enforcement. With ODL, you can maintain a simpler underlay. For cloud applications, you can co-manage virtual compute, storage, and networking hardware. 

Greenfield campus networks can be structured to be secure without the performance impacts that are common with traditional campus architectures. Using ODL, these networks can be based on an architecture explicitly designed to accommodate high-performance applications on a virtually dedicated science network that is distinct from the general-purpose network. ODL provides hardened isolation, dynamic instantiation of services, and fine-grained flow steering, all of which are crucial for increasing performance while meeting security demands.

Smart cities and other IoT applications may involve multiple device types (both fixed and mobile) and multiple different networking technologies including fiber optics, IP/MPLS, and LTE or 5G wireless networks. ODL can program all of these technologies on thousands of connected devices.

ODL supports the varied protocols and scale (such as NETCONF and PCEP for thousands of devices) needed for wide area research and education networks. ODL is uniquely suited to advanced applications such as bandwidth on demand and traffic engineering.

Examples

Bristol: A Programmable City Based on OpenDaylight

The Bristol is Open project collects information about many aspects of city life--such as energy, air quality and traffic flows, from a large number of sensors--including smartphones and sensors mounted on lamp posts.

Bristol's SDN solution, built with the ODL framework, manages M2M communications from/to the IoT devices, offering greater flexibility and agility than a hard-wired network infrastructure. Using ODL, devices attached to the network can be programmed “on the fly” by network managers as new requirements and services emerge.

Data is then accessed through a portal and presented in human-friendly form, such as maps and diagrams showing real-time pollution, journey times, energy efficiency, or the results of ad hoc polling for civic or entertainment matters. There is also a bandwidth-on-demand component to the ODL solution wherein capacity can be guaranteed to selected users--the BBC is currently taking advantage of this, donating research fees for high bandwidth access.

Cornell University: High Performance Campus Based on ODL

Cornell University is using OpenDaylight to put together a very high performance research network. Initially not knowing exactly what their applications would be, Cornell decided to go with OpenFlow for maximum flexibility, and implemented one of the largest pure OpenFlow networks ever deployed into production at a university. Sample initial projects include a “deep learning” project that uses network-intensive algorithms to automatically identify faces in video streams and catalog billions of objects in space.

The ODL-controlled network delivers nearly 40 terabits-per-second of bandwidth to over 8,000 students and faculty at CIS (supporting 10GE to the desktop) and has already--due to the ability to isolate research projects and tightly control application flows--expanded to provide service and support to several other schools and departments at Cornell.

GEANT: Bandwidth on Demand Service with DynPaC Framework

GEANT is introducing SDN capabilities in the backbone with a Bandwidth on Demand (BoD) service. This service uses their DynPaC Framework, which offers dynamic and adaptive traffic engineering using path computation elements. DynPaC provides efficient use of network capacity, resiliency for link failures with quick recovery times, reduction of the operational costs of service management, and improved network monitoring by gathering and immediately acting on real-time information.

Figure 1: GEANT's DynPaC Framework Implemented with OpenDaylight (Source: www.geant.org)

The DynPaC service manager acts as the coordinator, orchestrating the interactions between the other modules in the framework, and monitoring network events to determine how to best react to changing conditions. It introduces new services (moving flows to alternative paths if necessary), provides resiliency and fault recovery, and schedules services.

The Service Manager uses OpenDaylight’s Path Computation Element Protocol (PCEP), which obtains the physical topology and computes the shortest path between two network points, taking bandwidth constraints and scheduling information into consideration. The Service Manager and OpenDaylight then provide optimal path choices between source and destination nodes, prioritizing the routes to minimize the number of split flows.

GEANT is very active in the ODL community, contributing both original and maintenance code.