- Product
-
Solutions
Digital ExperienceCloud Adoption
- Customers
- Resources
- About
- Login
- Start Free Trial
Making Hybrid Cloud Perform
A modern retailer uses ThousandEyes to
understand and optimize application and
service performance over internal networks
and the Internet.
About the Company
This retailer has over a thousand employees and store locations across the United States, as well as an active e-commerce website.
Troubleshooting a Modern, Hybrid Cloud
This North American retailer is forging a hybrid cloud approach to its infrastructure and applications. Revenue-generating applications
such as order entry, web origin servers and core productivity applications such as Microsoft Exchange and VPN remote access are hosted in
two private data centers. The data centers are connected to the headquarters office and dozens of store locations via a carrier MPLS-based
WAN service. They also leverage cloud-based services such as a content delivery network (CDN) for its e-commerce website, as well as SaaS
providers for business-critical functions such as sales tax calculation, credit token authorization and payment processing.
The company's IT infrastructure needed fast, detailed visibility into the application impact of network issues. Their senior network engineer said, "In the past when I've had to deal with application troubleshooting, I've had to rely on a traffic analyzer. That means grabbing packet captures, running those through analyzers, and purchasing costly software suites or appliances to drill down into a waterfall view of what's happening. The purpose is to nd out where latency is occurring. Is it a client waiting for a server, or is it loss on the network? The problem is that analysis can potentially take a multi-day effort. That's just not an efficient way to operate."
The company's IT infrastructure needed fast, detailed visibility into the application impact of network issues. Their senior network engineer said, "In the past when I've had to deal with application troubleshooting, I've had to rely on a traffic analyzer. That means grabbing packet captures, running those through analyzers, and purchasing costly software suites or appliances to drill down into a waterfall view of what's happening. The purpose is to nd out where latency is occurring. Is it a client waiting for a server, or is it loss on the network? The problem is that analysis can potentially take a multi-day effort. That's just not an efficient way to operate."
"The combination of Cloud Agents and Enterprise Agents being able to execute tests and using the ThousandEyes platform to analyze performance and network data on both internal and external applications on a single screen is extremely appealing and powerful."
Senior Network Engineer
Monitoring Internal and External Applications
The network team deployed ThousandEyes to monitor both internal applications delivered over their MPLS-based WAN and external-facing applications delivered over the Internet using a combination of Enterprise Agents and Cloud Agents. "The combination of Cloud Agents and Enterprise Agents being able to execute tests and using the ThousandEyes platform to analyze performance and network data on both internal and external applications on a single screen is extremely appealing and powerful," commented the network engineer.The network team started by using Cloud Agents to monitor data center-hosted VPN and Outlook for Web (OWA) access from remote employees. They expanded their use of ThousandEyes, adding monitoring of retail applications accessed from retail locations, CDN-to-origin server performance, then further extended the deployment to cover SaaS services accessed from data center applications.
Figure 1: ThousandEyes Dashboard provides easy to understand visualizations on application and network performance.
Speeding Mean Time to Repair
The network team leverages ThousandEyes to pinpoint network issues impacting application performance without the time and costly capital investment of packet capture appliances and approaches.The network engineer explained, "With the way that application performance, network performance metrics and path visualization information is presented in the ThousandEyes dashboard, I can get those same session maps, those same waterfalls, and deep network insights without having to do any packet captures and feeding them through a traffic analysis tool. I don't have to have that always-on, packet capturing mentality because I know that based on these tests, at the rate that they're occurring, I'm going to have enough historical information. I know I'll be able to go back and pinpoint a single object within this single application that was churning for 300 milliseconds waiting for a response from a server and know if it's the application, the MPLS WAN, or the Internet. It's a much quicker and easier way to pinpoint and solve those application issues."
"With the way that application performance, network performance metrics and path visualization information is presented in the ThousandEyes dashboard, I can get those same session maps, those same waterfalls, and deep network insights without having to do any packet captures and feeding them through a traffic analysis tool."
Senior Network Engineer
Eliminating Finger-Pointing over SaaS Performance
ThousandEyes has also enabled the IT team to transform the nature of service escalations with their SaaS providers from finger-pointing to problem-solving."Previously, when we had service interruptions to things like our payment processing partner, usually our SAP team would get on a phone call, there would be rounds of discussions, and after a good amount of time, there would be a sort of negotiated conclusion," noted the engineer. "But it was always kind of nebulous. Was it our problem, was it their problem? Sometimes when you get involved with third-party service providers, there can be finger pointing, for lack of a better term. And it's tough to reach a resolution."
That dynamic changed after deploying ThousandEyes. Soon after, the team was able to identify a service interruption and share concrete data with the provider that showed conclusively that their service was unavailable from multiple locations.
"Going to them with that type information changes the tone of the service escalation," said the engineer. "It becomes less of an exploratory conversation: 'Well, it could be this, it could be that,' and becomes, 'Hey, we saw this, we witnessed that, and here's our supporting information.' In this case, we showed evidence that other service providers' traffic was going through the same network and Internet points without a problem, which helped illustrate that the root cause was in their service specifically. They then dug into the incident, came back and verified that in fact there was a service interruption that their operations team hadn't been aware of until we provided data from ThousandEyes."
Figure 2: Retail engineers use ThousandEyes Path Visualization to solve complex application delivery issues.
"We showed evidence that other service providers' traffic was going through the same network and Internet points without a problem, which helped illustrate that the root cause was in their service specifically."
Senior Network Engineer
Building Baselines for Smarter Alerting
Beyond troubleshooting and SaaS service escalations, the network team has started to build baselines of application performance and network behavior to identify anomalies better. The engineer said, "Where there's smoke, there's fire. If we know how things are supposed to look, and they start to look a little bit different, we can pay proper attention to them. The problem is understanding what is normal. Without that understanding, you can't detect problems. Also, when you inevitably run into performance troubleshooting, it doesn't make sense to try to get performance back to a state that never existed in the first place.""Previously, we wouldn't have had a way to build a picture of what is normal," added the network engineer. "But with ThousandEyes, based on application performance trending data that we're collecting, we have a baseline. So now we know that if performance varies from the 90th percentile baseline, we can start a productive investigation into a potential issue."