“Bandwidth patterns in cloud environments are constantly changing, creating dynamic bandwidth bottlenecks in static transport networks. SDN and related software finally give us the tools and all-layer access to write applications to create agile and flexible data flows,” said Robert Cannistra, senior professional lecturer, Computer Science and Information Technology, Marist College. “The only way to meet the fierce bandwidth demand while improving the end user experience is to make the network more dynamic and responsive. With the technology in our test bed, you only need to setup the network once. Afterwards, everything else is handled automatically in response to real-time performance metrics.”
In the SDN demonstration, three geographically diverse data centers with switching, server and storage technology from IBM are interconnected by a fully reconfigurable and OpenFlow-enabled optical network based upon the ADVA FSP 3000. An open-source SDN application, developed by Marist College, dynamically provisions the optical circuits and monitors, manages and manipulates (create, modify, delete) end-to-end flows across all layers of the network.
In the first stage of the project, a web interface running on an open sourced OpenFlow controller was developed to enable on-the-fly provisioning as well as downloading of pre-defined traffic profiles to quickly reconfigure optical traffic patterns. In a second stage, the OpenFlow-enabled Layer 2 switching fabric from IBM was added to the network highlighting full network virtualization across both Layers 1 and 2. In the final stage, the fully automated movement of Virtual Machines (VMs) between data centers was achieved in response to alarms triggered by VM monitoring software, including the provisioning of completely new optical circuits, thus demonstrating how relevant data center applications can benefit from a virtualized network approach.
Cloud computing and data center virtualization are creating a more volatile environment than ever before. VM loads are constantly being balanced across geographically diverse clusters and locations to use compute resources more efficiently. It is no longer economically viable or physically feasible to overprovision transport networks for peak demand. The answer is flexibility and scalability on all network layers enabling an end-to-end, all-layer provisioning of flows.
“This demonstration reveals how resources can be cost-effectively pooled among data centers by automating wavelength provisioning, commissioning and assignment, while equipping the network to dynamically respond to on-the-fly changes in application workload and traffic patterns,” commented Casimer DeCusatis, distinguished engineer, IBM.
“When end-to-end provisioning of an “application” aware network becomes automated and therefore instantaneous, a whole new paradigm in network orchestration arrives,” said Todd Bundy, business development director, ADVA Optical Networking. “In addition to load balancing across data centers, dynamic functions such as bandwidth calendaring, “follow-the-sun” applications and new more cost effective disaster recovery mechanisms become possible as well. I can now push a button and instantly accomplish what used to take dozens of people across multiple departments weeks to months to do. The possibilities are profound.”