BT: Telco Mindset Could Negate Container Benefits

PARIS -- MPLS, SDN and NFV World Congress -- It's widely recognized that using lightweight containers as an alternative to more clunky virtual machines could lower costs and open up new service opportunities for telcos investing in SDN and NFV technology. But if they are to really benefit from containers, operators will have to start treating their network functions less like pets and more like cattle, says Peter Willis, the chief researcher for converged networks at the UK's BT. (See Containers Key to Virtualization.)

"If a function becomes ill today we spend a lot of time diagnosing it and caring for it and bringing it back to health," Willis told attendees at today's MPLS, SDN and NFV World Congress in Paris. "To take advantage of containers, you have to move to a different model and treat them like cattle -- if one gets ill you shoot it and start again."

Shooting a big cow on which a lot of customers depend doesn’t sound very advisable. But if each cow was very small -- in fact, if you had one cow per customer -- the sacrifice would become much easier to perform and carry far less risk, says Willis.

What that really means for service providers is an entirely different operating model based on the concept of microservices, whereby the containers that operators use are pared down to the bare essentials.

The danger, explains Willis, is that operators end up "putting everything in the container they had in the original monolithic virtual machine." They might be able to realize some efficiency improvements, but the gains would hardly be significant.

"A microservices model is when you only implement the function you need and not the kitchen sink," says Willis. "This is the whole world of DevOps, where if something is broken you don't try to address the failure directly but go back and fix the code. It's a mindset change for telcos and they have a long way to go."

Changing this mindset is possibly the biggest challenge that telcos will face in adopting containers, but it is by no means the only one, according to the BT Group plc (NYSE: BT; London: BTA) executive.

Another concern is the lack of container management systems that can support multiple network interfaces natively. Encouragingly, Willis says there has recently been some progress in this area, referring to a technology called Multus that appeared in December 2016. "That is worth looking into," he says.

Security remains another headache, according to Willis, but the real issue is being able to demonstrate the business benefits of containers to the telecom community. "We need to qualify those to encourage vendors to create VNFs and operators to address their mindset challenges," says Willis.

This is not the first time the BT researcher has come up with a list of challenges for those developing or investing in SDN and NFV technologies. Clearly not afraid to speak his mind, Willis has previously drawn attention to the perceived shortcomings of the OpenStack open-source platform from a telco perspective. He's even suggested that BT might ditch OpenStack and instead use a proprietary technology unless the problems are addressed. (See BT Says OpenStack Still Not Up to Spec, OpenStack Doubts Surface After BT Ultimatum and BT Threatens to Ditch OpenStack.)

For more NFV-related coverage and insights, check out our dedicated NFV content channel here on Light Reading.

On the subject of containers, though, he is effusive about the benefits they could have in a variety of use cases.

Containers might be a particularly efficient way of doing the "network slicing" that gets talked about so much in the context of 5G. Essentially, this would mean dividing network resources into different "slices" that cater to the needs of specific vertical-market customers.

"With virtual machines you get a lumpy and slow-to-respond system, but with containers you could adapt more quickly," says Willis.

Using containers, operators could also reboot systems far more speedily than with virtual machines. That means they could start to think about phasing out some of the equipment they maintain for backup purposes.

Containers might also have a role to play in the development of edge or fog computing, whereby IT resources are moved from centralized facilities to base stations or aggregation points closer to end users. To process information locally, Willis explains, operators might need to load functions onto the edge devices themselves.

"With such lightweight devices you can't afford a full virtual machine that might take up a couple of gigabytes of RAM [random access memory]," he says. "But a lightweight container could do that very easily and we've been experimenting with this on IoT [Internet of Things] gateways."

— Iain Morris, Circle me on Google+ Follow me on TwitterVisit my LinkedIn profile, News Editor, Light Reading

clarkede 4/1/2017 | 10:21:03 AM
Valuable insights from an NFV pioneer Peter Willis has pioneered making NFV fit for purpose at BT. He coined the term "NFV" in a telco meeting in Paris in 2012 where he also presented the industry's first results on NFV performance and TCO. These rare statements from Peter are always insightful and based on real implementation experience. For those of us driving NFV standards and open source such insights coming from a credible source are of high value in showing where our focus should be.
Sign In