AT&T, Ericsson virtualise IMS CSCF on OpenStack

AT&T and Ericsson announced that the companies have jointly demonstrated a virtualised IMS (IP multimedia subsystem) node call session control function (CSCF) on OpenStack using Ericsson’s cloud execution and management platform, according to BCN.

The CSCF is a key component of converged network infrastructure. It controls the fixed and mobile IP multimedia subsystem and manages all the signalling from end-user to services and other networks.

The companies showed how a CSCF virtual machine can be migrated from faulty hardware to a working backup without losing any critical data.

“AT&T and Ericsson have been working together on a prototype to run a virtual network function on OpenStack,” Toby Ford, AVP cloud technology, strategy and planning at AT&T. “In addition to actually setting up a multimedia session and the demo system shows some of the high availability features needed for the telco applications.”

The companies demonstrated the CSCF virtual machine being provisioned on OpenStack via Ericsson’s Cloud Execution Environment Platform, with the company’s Cloud Manager orchestrating and managing the virtual network function.

Mats Karlsson, head of architecture and process, group function technology at Ericsson said that at the moment, only a very limited set of network functions have been virtualised.

“We plan to virtualise the majority of Ericsson’s network functions by 2016. This is an important step towards the virtualisation of telecom network functions as the demonstration shows critical function like IMS can be virtualised without jeopardising high availability requirement telecom services demand,” Karlsson said.

Ericsson has been busy bolstering NFV for operators as of late. The company recently helped found Open Platform for NFV (OPNFV), an open source project started by vendors, cloud providers and telcos aimed at developing a carrier-grade reference architecture for virtualised network applications, and it has also launched two labs aimed at certifying vendors and telcos using the architecture.

Steven Peacock, head of cloud services, business unit global services at the Swedish networking vendor told BCN that most of the problems associated with virtualising telco applications have to do with the complexity of the service chains behind the scenes.

“To some extent it’s easier to virtualise many applications in an enterprise environment because you don’t have such complex integration required. In the telecoms space service chains are highly integrated; each service in the chain comes with a specific set of policies that then has to be able to talk to the next service. In a distributed environment where hardware integration is still key, that can become very challenging,” he explained.

“These policies are currently embedded within the applications but there’s definitely discussion around whether to centralise them in a service catalogue kind of way,” he added.

That said, a common NFV architecture that works in a multi-vendor environment could go some way towards solving virtualised function integration, and Peacock said that as SDN-enabled hardware becomes more prominent some of these challenges are likely to subside.