IBM, AT&T and Applied Communications Sciences have developed a technology that could reduce the time needed for networking cloud environments to seconds, creating elastic bandwidth capabilities between data centers to handle spikes in traffic.
The proof-of-concept, unveiled by the three companies July 29, was developed as part of a 7-year-old program of the U.S. Defense Advanced Research Projects Agency (DARPA) aimed at creating technology for both the government and commercial sectors that allows for the rapid reconfiguration of massive terabit networks. The technology, part of DARPA’s Dynamic Multi-Terabit Core Optical Networks—or Coronet—program, is based on software-defined networking (SDN) principles and leverages OpenStack, the open-source cloud orchestration software.
The rise of cloud computing has created a demand for greater automation and lower costs for IT infrastructures in increasingly massive data centers. While such automation and dynamic provisioning is becoming more commonplace in servers and storage, networking within and between clouds continues to be a mostly manual operation that is time-consuming and expensive, according to Robert Doverspike, executive director of network evolution research at AT&T Labs.
The proof-of-concept technology could change that, Doverspike said in a statement.
“These shifts have driven the need to develop rapid and high rate bandwidth-on-demand in the wide-area network,” he said. “By combining software-defined networking concepts with advanced, cost-efficient network routing in a realistic carrier network environment, we have successfully demonstrated how to address this need.”
The result could be greater efficiencies and lower costs for cloud service providers and carriers, the companies said. It also will enable organizations to move massive amounts of data between cloud data centers in the event of an emergency or disaster.
AT&T is in the midst of what officials are calling Domain 2.0, a wide-ranging transformation of its network that leverages SDN and network-functions virtualization (NFV) technologies to make the infrastructure more manageable, scalable, automated and cost-effective by taking many of the network tasks—such as policy control, load balancing and firewalls—out of expensive networking hardware and putting them into software that can run on commodity servers.
AT&T is following in the footsteps of such massive Web-based businesses as Facebook and Google, which have developed their own networking technologies in-house.
The proof-of-concept demonstration leveraged AT&T’s SDN WAN Orchestrator to route server connection requests to the right network layer, while Applied Communications Sciences (ACS) created provisioning protocols that were integrated with commercial transport Dense Wavelength Division Multiplexing (DWDM) network components to enable the technology to create and tear down connections as needed by the workloads and networking traffic.
IBM brought its cloud platform and orchestration technology to manage the life cycle of virtual machine network applications on OpenStack software. The IBM contribution enabled the software to automatically monitor the workload on servers and request network bandwidth from AT&T’s SDN WAN Orchestrator and compute resources needed to migrate the VMs from one cloud to another.
According to the companies, provisioning time for the networks was as fast as 40 seconds, with the possibility of hitting times of less than a second in the future.
“This technology not only represents a new ability to scale big data workloads and cloud computing resources in a single environment but the elastic bandwidth model removes the inefficiency in consumption versus cost for cloud-to-cloud connectivity,” Douglas Freimuth, IBM Research senior technical staff member and Master Inventor, said in a statement.