Enterprise lukewarm to SDNs
24 June 2015 | 0
Even though service providers are well on their way with software defined networks (SDN), enterprise adoption of the technology is slowed by a host of issues.
Chief among them is cultural inertia. Large enterprises in particular are loathe to change anything, be it technology, operational processes or organisational structure, especially if the need to do so is unclear or viewed as potentially risky.
Another is the business case. Implementing open source software on bare metal hardware — consistently referred to at the Open Network Summit conference as the building blocks for open, vendor-independent SDNs — may initially reduce capital cost but the long term benefits of operational cost reduction, ease of use, automation, flexibility, high performance and orchestration may be difficult to justify or align with a company’s core business.
Integration with an existing, or brownfield, environment might be another discouragement. Businesses could be entrenched for decades in a particular brand or way of networking, and have an investment of tens of millions of dollars in infrastructure and trained staff to operate it. Trying to integrate SDN technologies, be they open source or vendor sourced, could prove disruptive, costly and time consuming, especially for enterprises that lack the integration skills to do so.
The last impediment is the immaturity of SDN technology. Though they have made progress as far as stability in the last five years since the ONS conference started, SDN offerings are still largely unproven and battle tested in mission critical large enterprise environments.
“If one person does it in a compelling way, everyone will follow,” says Guru Parulkar, executive director of the Open Networking Research Centre and ONS. “But they need one or two early adopters to demonstrate it.”
The US National Security Agency is reaping the benefits of a small scale OpenFlow SDN in its enterprise IT environment. The agency is using SDN to gain greater control and visibility over its network, for security and more efficient operation. But even deploying the SDN in a small part of its campus network, and then attempting to expand it, was bogged down in bureaucracy and cultural resistance.
“This is actually a really big problem,” says Bryan Larish, NSA technical director for enterprise connectivity and specialised IT services. “The technology, quite frankly, is the easy part. It’s how do we change the culture, how do we affect this massive machinery to make a move in a new direction.”
“It’s not simple when it completely changes how a company traditionally operates,” says Robert Bauer, IT director at Asurion, a provider of device protection services for smart phones, tablets, consumer electronics, appliances, satellite receivers and jewellery. “The people and culture changes, that’s what kills you.”
“In a large enterprise, change is the enemy,” says Vesko Pehlivanov, vice president of technology services and strategy at Credit Suisse, who also spoke at ONS.
With 5,000 to 6,000 business applications, Pehlivanov has to present Credit Suisse with a compelling business case for moving to SDN and perhaps injecting new software code, including open source, into that environment. The benefits of open source SDNs to cloud and service providers were prevalent at ONS this week but not so apparent to a risk averse, highly regulated investment bank.
“Is software the right model for enterprises as well?” Pehlivanov asked during his ONS keynote address.
Even if there was a fit for open source in the Credit Suisse SDN — Pehlivanov said it was critical for attracting the next generation of IT talent — the bank could not reciprocate.
“Most of our intellectual property is from code we developed,” he said. “We cannot contribute back to the open source community.”
And disruption is another issue. Introducing change into a legacy environment with internally developed code and 5,000 to 6,000 business applications while still maintaining reliability is a daunting proposal.
“How do you handle entropy in the environment?” Pehlivanov asks. “How do we guarantee stability?”
Technology immaturity doesn’t help. Advances are being made in fortifying open source and vendor developed SDN and cloud code every year, but melding them into brownfield environments where everything is working and predictable may not be appealing.
“It would help enormously” if vendors underwrite open source projects like OpenStack to indemnify users if something blows up, says Colin Constable, co-founder of Deutsche Bank Labs, a research and development arm of Deutsche Bank.
“I run OpenStack,” Constable says. “It runs ok for a week and then something goes wrong I’m like, I am not an expert on OpenStack. You find out pretty quickly that you’re not an expert on OpenStack when OpenStack goes wrong. The only way I can fix it is by re-installing the whole damn thing which means, there’s no way it’s going into production unless I’ve got somebody like Red Hat, unless I’ve got somebody like SUSE to provide that expertise on demand. This is complicated stuff.”
Which is why enterprises are looking at someone else to take the arrows as an early adopter and prove it out before they do.
“We’re looking for the industry to help us solve these challenges,” says Credit Suisse’s Pehlivanov.
“The cloud providers are doing amazing things but others are waiting to do it,” says Parulkar. “There is a gap. The rest of the industry seems very slow.”
Jim Duffy, IDG News Service