GÉANT Testbeds Service - Bringing researchers and technology together


The pace of research into advanced networking technologies and services is accelerating with new systems being developed to take advantage of expanded network capabilities. This research is not just limited to developing new networking technologies but is increasingly looking at operating new services over these networks.

Initially research can be done in simulation or on the lab bench but soon it is necessary to test these models in real-life systems requiring research on high-performance networks and equipment. This leap from bench to real-world can require significant technical expertise, time, administrative and logistical support, and funding. Additionally, these research teams themselves may not have the skills and experience to build out wide area networks adding to cost of additional team resources and increasing the ramp-up time required prior to doing the actual intended research.

Virtualisation - No longer a poor substitute

Over the last two decades there have been many projects available across the world, trying to make networking facilities available to researchers. Initially emulation was an intermediate solution trying to simulate the equipment while protecting the production environment from tests. This provided an option of creating the networking environment researchers did not have funds or resources to create, but these emulations were unable to adequately replicate real-life performance of real networks, making it hard to assess performance of systems when at the limits of technology.

With this in mind, the GÉANT Testbeds Service (GTS) was conceived to address a specific purpose –advanced network research, at scale. GTS is the fastest and most cost effective way to field a wide area network testbed to validate models in a real-life environment. Most importantly with GTS we are mitigating the setup process with a simple method of defining and getting the needed resources within minutes. GTS offers an easy, agile procedure for providing researchers with secure and immediate access to predefined sets of virtual resources. This allows researchers to build, test, break down and rebuild networks incredibly quickly.

Virtualisation is not "emulation" or "simulation" and does not mean slower than real-life. Today's virtual objects are predominantly realised in real hardware. Virtual machines, virtual circuits, virtual file systems are all examples of production-capable high performance virtual resources accessible within the GTS environments. Combined with advanced virtual SDN switching, access to dedicated computational resources, intuitive user interfaces, and a broad range of other usability and control features, GTS delivers unique capability to define custom virtual environments deployed across Europe to run exciting new services and applications. With GTS, virtualised resources can provide elastic WAN networks that leverage network infrastructure effectively and efficiently and can deliver state of the art performance.

In addition GTS is able to provide some advanced features not available in other virtual environments. One key design objective of GTS environments is that they are isolated from one another or the real world. This enables researchers to run parallel tests, sharing common hardware platforms, without interfering with each other. This model of operation provides facilities in the wide area network that enables researchers to deploy novel concepts at scale without the fear that an errant experiment could impact other production services. This is novel. And it allows service providers to support such highly experimental activities in the WAN precisely because there is secure isolation among GTS testbeds. This allows novel services to be exposed to real users, to evolve in realistic service environments, and be deployed months or years sooner than traditional methods.

The GTS team is currently preparing to transition from GTS version 3 to GTS version 4. As part of this evolution a number of important new features will be introduced: dedicated high performance server resources will arrive, support for full 10Gbps circuits, virtualised high performance SDN switching resources, and multi-domain international reach will be introduced.

Additional features are on the GTS roadmap into 2018. Users will be able to modify their environments in real time, and save those environments and restart them later. A wider range of transport circuits and SDN switching and forwarding capabilities will be available. Future GTS environments will be able to run 100 GBPS experiments with full packet processing performance, federated AAI and advanced user policy control features will be available.

More than just a network environment

GTS is expanding its abilities to support more and different types of research. GTS can now build 'application specific' networks that bind production science communities, provide a rapid deployment "on ramp" to operate and refine new services, or act as an advanced learning platform for computer science and network engineering curriculum. This moving up above the traditional network technologies to support the next generation application and service development will help developers test their services in real-world scenarios that can be set up and controlled via a website in an automated process.

This automated provisioning allows researchers to drive faster test and improvement cycles which speed up innovation in science and technology. For instance, particular GTS environments could be used to support security services and DDOS mitigation tools in a production scale environment, or to support novel big data and information centric networking paradigms.

To find out more about how GTS can support large scale network service research and development visit http://www.geant.org/Services/Connectivity_and_network/gts

This ​article appeared in 

CONNECT Issue #24​​​​​​​​