MOUNTAIN VIEW, Calif. -- Open Networking User Group Spring 2016 -- Network monitoring tools aren't up to the job of delivering the reliability needed for applications spanning multiple clouds, according to enterprise network operators.

For example, applications spanning Oracle, SAP, Microsoft, Amazon clouds and the public Internet will emerge in less than six months, and the industry lacks the ability to monitor those kinds of systems, said Joe Ferrell, leader and CTO, computing and datacenter infrastructure, GE.

"If operational teams don't have instrumentation and skillset to troubleshoot, it can be really frightening," he said.

Figure 1: On Edge Stanford University Professor David Cheriton (far left) is on the edge of his seat, listening to GE's Joe Ferrell (far right). Also in photo (l-r): UBS's James Younan, BNY Mellon's Neal Secher, Intuit's Pablo Espinosa, and FedEx's Aryo Kresnadi.

James Younan, UBS director, infrastructure research and development, said real-time network monitoring is essential for network innovation. "I need to be able to consume events in real time and be able to tell you what the state of the platform is," he said.

Academic and government networks have sensors that emit small amounts of data that can be consumed in real time, he said. Hadoop and Kafka can build an analytics clusters for a monitoring network.

SNMP (Simple Network Management Protocol) isn't up to the task. "I get the calls from the user before I get the SNMP trap that something went awry," Ferrell said.

Network monitoring tools, which can consume a quarter of network bandwidth, need to get leaner, said one ONUG attendee speaking from the audience.

But David Cheriton, a Stanford University computer science professor and venture investor, disagreed. "Why is 25% an unreasonable price to stay alive?" he said.

Find out more about enterprise cloud at our upcoming Big Communications Event in Austin, Texas, May 24-25. Register now!

Cheriton described a conversation with an engineer at Audi, who said 90% of the software in their vehicles is diagnostic. "Is that wrong? Maybe not. Maybe we should adjust our thinking that this is an unreasonable price to pay. This is a matter of survival," Cheriton said.

However, when monitoring traffic becomes a drag on the network, that's a problem, Neal Secher, managing director and head of network architecture at BNY Mellon, said. "When the monitoring steps on somebody's job, an actual application process, they'll say, 'Whoa, whoa, the doctor is killing the patient! You have to turn this off,' " he said.

Network operators can have the best of both worlds by doing processing at the edge, discarding inessential information and only bringing the rest to the center of the network, Secher said.

— Mitch Wagner, , West Coast Bureau Chief, Light Reading.