AT&T has taken its disaggregation efforts to the converged IP edge by working with Broadcom, UfiSpace and Cisco. The telco had previously teased the use of white boxes at the edge before announcing on Monday that the open, disaggregated IP edge routing platform was now live in its production environment.
According to an AT&T spokesperson, the telco has had production traffic on its open disaggregated solution for about a month with no reported anomalies.
RELATED: AT&T's Fuetsch says AT&T now has internet white boxes in production
AT&T said the first IP edge use case is peering, which is the system that connects AT&T's IP network with other internet service providers, according to a statement by Andre Fuetsch, AT&T’s CTO for network services. Fuetsch said AT&T was also working with Cisco to develop additional edge use cases such as broadband, IP content, Ethernet, mobility and VPN services
"We have started the journey to converge the disparate edge implementations we have today onto common software and hardware driving uniformity, simplification, and agility," Fuetsch noted.
AT&T is using white box routers from UfiSpace, which it's also using for the previously announced IP/MPLS core deployment. Over a year ago, AT&T put its specifications for its distributed disaggregated chassis (DDC) white box architecture into the Open Compute Project (OCP.)
RELATED: Spotlight is on DriveNets with AT&T's deployment
In September, AT&T acknowledged for the first time that DriveNets was providing the core-networking routing software for its next-gen core network. It's noteworthy that Cisco got its foot in the door for the converged IP edge routing platform by providing its IOS-XR software.
RELATED: Raynovich—Is Cisco trapped? AT&T/DriveNets calls Silicon One into question
Cisco’s IOS-XR network operating system (NOS) provides the management and control functions for AT&T's IP edge. Some of the key features for the peering use case are BGP Flowspec, unicast Reverse Path Forwarding (uRPF), and port mirroring that collectively enable granular monitoring and mitigation actions for security threats such as directed denial of service (DDoS) attacks.
"With this announcement, AT&T has demonstrated its commitment to disaggregated designs, from the IP/MPLS core to their IP edge, with a nod to future use case expansion," said Roy Chua, founder and principal at AvidThink. "The platform is the same DDC previously mentioned in their announcements with startup DriveNets for their IP/MPLS core. What's interesting is Cisco's almost-willingness to move towards disaggregation—from silicon to hardware to software. CSPs and cloud providers are applying pressure directly by demanding compatibility with open-source initiatives from Open Compute Project.
"They are also indirectly pressuring Cisco by working with a new breed of network software players like DriveNets, Arrcus, Pureport, and Volta Networks. I find it notable that Cisco's IOS-XR with DDC is being used at the edge and for peering, while DriveNets' software was deployed in the core, albeit not across all locations. An interesting reversal of roles and not something I would have expected even five years ago."
AT&T's DDC design, which was built on Broadcom's powerful Jericho2 family of merchant chips, aims to define a standard set of configurable building blocks on less costly service-class routers ranging from single-line card systems, known as "pizza boxes," to large, disaggregated chassis clusters.
With the IP edge announcement, AT&T has deployed the Jericho2 DDC design, which has also been accepted into OCP, in both its network edge and the core routers that make up its global IP common backbone.
AT&T is also employing the same white box hardware from UfiSpace for both the disaggregated core and edge router deployments. The UfiSpace portfolio consists of three components: a 10x400 GigE Line card system based on Broadcom’s Jericho2 packet forwarding engine, a 10x400G line card system that also uses Jericho2, and a 48x400G fabric system based on Broadcom’s Ramon chipset. Those building blocks are deployed on a modular architecture based on site needs using the DDC design in various configurations to build routers with capacity anywhere from 4 Tbps to 192 Tbps.
“This is a really big development in the networking ecosystem,” Fuetsch said in the press release. “This model gives us options and flexibility in our supply chain and enables us to use best-in-breed products whether they come from established or disruptive suppliers. And this is well past lab experiments; the technologies and ecosystem have matured, and we are now into the production deployment phase.”