By capturing and analyzing packets with Riverbed AppResponse, OneMain Financial quickly pins slow performance to the network or application. The right team gets to work sooner, aiming to fix problems before users notice.

Challenge: delivering a great application experience

Quick decisions are a competitive advantage for lenders—and OneMain Financial aims to make an offer just 20 seconds after a customer applies online. Slow credit checking or electronic signatures, for instance, can’t be tolerated. “If a page doesn’t load or the progress bar stalls, the customer might give up and try another lender,” says Richard Hurst, supervisor of network services for OneMain Financial.

Troubleshooting is tricky in OneMain’s complex environment. The company’s software-defined network connects 1,600 branches, 12 campuses, the three major credit-reporting agencies, and public clouds used for customer-facing websites and extra capacity during busy times. “When application performance slows, it can become a blame game,” Hurst says. “The Internet service provider, credit agency, and other parties tend to assume the problem is someone else’s.”

OneMain Financial’s existing diagnostics tools were outdated. Hurst wanted a modern packet-capture tool that made it easy to see whether performance problems originated in the network or application.

Solution: capturing and filtering packets to find the evidence

OneMain Financial found its answer in Riverbed® AppResponse, a Network Performance Management (NPM) solution Hurst had used at another company. AppResponse includes a license for Riverbed Packet Analyzer Plus, which filters packets based on variables like time and network interface. “The combination of AppResponse and Packet Analyzer Plus makes it easy to find the precise set of packets I need to see if a performance problem starts with the network or the application,” Hurst says.

Before OneMain started using AppResponse, application owners who needed packet data submitted a trouble ticket. “Since we started using AppResponse, ticket volume is way down,” Hurst says.  “Instead of opening a trouble ticket, application owners just shoot me an email asking for packet data for certain time period. I can get them what they need in a minute or two—compared to 2-3 days before AppResponse.” 

Benefits: faster issue resolution and a better customer experience

Some of the ways AppResponse has kept the business running smoothly:

Convincing a credit agency its application was slow. Over the weekend an application team told Hurst that one of the three credit agencies was taking longer than usual to return applicants’ credit scores and histories. The agency initially insisted its application was fine—pinning the blame on the network. But by filtering the weekend’s packets by time period, Hurst presented proof that the application was running six seconds slower than usual. Persuaded by the data, the agency fixed the problem.

Giving a vendor the data needed for troubleshooting. When OneMain’s server team reported slow server performance to its Linux vendor, the vendor requested 10 minutes of packet data—roughly 30GB. With AppResponse, OneMain quickly filtered out those packets and sent them as a single 30GB file. “Our other tools limit file size,” Hurst explains. “Without AppResponse, we would have had to send more than 100 files. AppResponse reduced a 2-hour chore to just five seconds.” After receiving the file from AppResponse, the vendor quickly resolved the issue.

Isolating the cause of slow electronic-signature application performance. Multiplied by a million transactions, even a 10-millisecond slowdown in the e-signature application can interrupt OneMain’s business. “With AppResponse, we can immediately see whether an emerging problem originates in the network or the application—even if the problem just started,” says Hurst. “With other tools we can’t pinpoint the source unless the problem takes place at the same time for a number of days.”

Microburst analysis. At one point Hurst suspected that certain network devices were causing bottlenecks, and was puzzled when another network-monitoring tool reported device utilization at just 50%. Hurst checked again with AppResponse, which analyzes packets in 1-millisecond intervals—not 1-minute or 5-minute intervals like other tools. “AppResponse revealed what other tools could not—that capacity sometimes maxed out for 15 seconds or even one second,” says Hurst. Upgrading the devices eliminated the bottlenecks.

Isolating the cause of voice quality issues. OneMain Financial’s voice network is generally very stable. On the rare occasions when latency affects quality, AppResponse quickly narrows down the problem to the service provider or the router so the right team can get to work.

Next steps: monitoring cloud and virtual environments

OneMain is now conducting a proof of concept using virtual and cloud versions of AppResponse. The company is also considering using Riverbed NetIM to capture network device metrics in the cloud.

Hurst concludes, “More and more teams are requesting data from AppResponse to figure out the source of application-performance issues. They’re grateful for an easy way to examine packets, and we’ve sped up mean time to repair and eliminated the blame game.”