Network testing in an IP world: No longer 'deploy, watch it blow up, respond'

No one could have imagined 24 months ago the volume and breadth of applications that would be running over operator networks today. Nor could we imagine all the hoops operators must jump through to make sure applications perform as they should, scaling to meet customer expectations—all while managing security and privacy concerns.

Today, intelligence in networks is no longer a “port-level-only” concern. Throughput is just part of the picture now, as operators have to be able to look at applications in a granular “content-aware” or “application-aware” fashion that zeroes in on applications and the policies that have to be applied to ensure things like QoS, entitlements or blockages of certain applications.

To help operators better test such IP application scenarios, Mu Dynamics has created a “crowd-sourced” packet capture repository that network testers can use to recreate real network scenarios (as the captured packets are based on actual network traffic). This week Mu announced its “pcaps” resource had reached 60 million packets captured, spanning more than 400 protocols. Those packets have been contributed and used by companies including
AT&T, France Telecom and Verizon Wireless, as well as network OEMs like Cisco Systems, Ericsson-Redback, Alcatel-Lucent, F5 Networks, Huawei Technologies, Juniper Networks, McAfee and Motorola.

The breadth and scope of the thousands of IP applications that those millions of captured packets represent means that operators must test and validate networks to make sure their policies are working at the same pace at which the applications are being introduced. There’s no longer the luxury of waiting six months to run tests, as was the case with legacy test tools.

“The ‘waterfall model’ of test and development has to become more agile when you have so many new applications emerging every day. With applications, you have to move capabilities to the customer premise rather than rely on your equipment vendor,” said Dave Kresse, Mu Dynamics CEO.

So for an operator to recreate a Facebook flow, or to test something as soon as it comes out on its network, more has to be done than just blast bits through a network. Rather, operators have to really grasp how their networks are handling applications running in real time.

That is difficult with traditional testing tools that fail to keep up with the rate of change of today’s environments. Most are throughput- rather than application-focused, which means identifying or recreating scenarios at an application layer is not possible without a lot of engineering to recreate a single application statefully. The end result is usually a close facsimile of applications.

“It’s not a true application payload that you are testing with older tools, so you really aren’t ‘exercising’ the network to know if policies will function and scale appropriately,” said Kresse.

His belief is operators need to use real traffic and they can’t take so long that the applications are obsolete by the time the tests are finally run. “To test and validate applications from a functional, security and reliability point of view, you have to take actual production traffic and transform it into test cases, and do so within minutes, not months.”

Additionally, operators have to become less reactive and more “real-time” in how they test what is happening on their network; after all, the ability to test traffic as it comes onto the network will directly tie to the customer experience.

“Tier 1 and Tier 2 operators know the value of testing because they are under tremendous pressure to stop attrition with a better customer experience. Deploy, watch it blow up then respond is not a good strategy in today’s application-driven market,” added Kresse.