How NXP Optimizes App Performance with ThousandEyes

This month saw our annual ThousandEyes Connect event return to the Brewery in London and this year we had the pleasure of hearing from Henk Jelink and Maarten Van Den Berg from NXP.

For those of you who aren’t familiar with the brand, NXP Semiconductors N.V. is a world-leading Dutch global semiconductor manufacturer headquartered in Eindhoven, the Netherlands, which employs approximately 31,000 people in more than 35 countries. Henk commented in his presentation, “everybody who has a car has at least 20, 25, or 30 NXP chips in it, you don’t know it but it really is the case. And everybody who has a mobile device has an NXP chip in it, or at least in 99% of cases.”

Henk has worked at NXP since 2000 and sits in their global connectivity group, whereas Maarten works for CGI and has been working as a contractor for NXP for more than 5 years as a connectivity architect.

Total Quality, Zero Defects, First Time Right

In his presentation, Henk mentioned that NXP has been working with ThousandEyes for the last 2 years. “It all started when a colleague stumbled on ThousandEyes on the internet looking for more tooling to do some research on some projects we had.”

“We had a lot of tooling for internal networks but as soon as you go on to the internet or towards the cloud you are blind, but I’m happy to say we’re not blind anymore”.

Henk continued, “The original driver came from Office 365. When Skype is not performing everybody is complaining. We all have that annoying colleague working in application support who blames the network and we know it’s probably not the network but you have to prove it. Finally we have the tools to really have good overview from the application level down to the network level and to troubleshoot those applications if needed.”

“We’ve made a lot of progress. The application support team are now convinced that they can use ThousandEyes to do troubleshooting on their own so they all now access the platform. Now they have a look at the platform before they start complaining which is nice!”

Since they originally adopted ThousandEyes, NXP have now grown to a stage where they’re leveraging 50 Enterprise and Cloud Agents to help improve the user experience for their workforce around the world. They are also leveraging ThousandEyes Endpoint Agents, Henk comments, “If an Enterprise agent shows no issue we will install End-point Agents onto the laptop if required to do a deep dive.”

NXP are also leveraging ThousandEyes native integrations for workflow management, “We also did an integration with ServiceNow with ThousandEyes alerting and are creating automatic tickets into ServiceNow for our support guys.”

Furthermore, Henk demonstrated how NXP has created a connectivity portal for its internal customers, enabling them to self serve and access key performance metrics for services they’re accessing, Henk added, “ThousandEyes is the most important tool in the portal”.

Key Use Cases for NXP

Henk was then joined on stage by Maarten, who was able to delve into some specific use cases on how NXP had used ThousandEyes to optimize particular applications for their global employees.

Early Visibility Gains, Even During the Pilot

Maarten began by explaining how ThousandEyes had helped them with some performance challenges from early on, even during the pilot phase. “We now have quite a large installed base with ThousandEyes Enterprise Agents in NXP, predominantly in Europe and Asia, and we’re starting to expand into the US.

For the pilot, we had an issue in our secure environment. We have a factory in Bangkok with a certificate server in Hamburg, so the connectivity needs to be good. Back in October 2017, when the pilot began, we were experiencing quite a lot of issues. We deployed ThousandEyes and discovered multiple issues straightaway that helped us solve those issues.”

“Another example was between Hamburg and our site in Kaohsiung. We used ThousandEyes to identify that we had some huge packet loss issues. When we looked into this we saw that the Hamburg LAN was causing the issue. ThousandEyes doesn’t necessarily show you exactly what is wrong but it does show you where it is wrong, which is really important in troubleshooting, otherwise where do you start?”

“ThousandEyes helps isolate the domain where you need to look for the underlying fault. If we had Device Layer from ThousandEyes at that time, we would have been able to click down and see more detail around the LAN. As it was, we discovered that we had a very overloaded core switch, which was causing the issue.”

Packet Loss with App Not Reachable

“Another example was when we were integrating the NXP and FreeScale networks, NXP having acquired FreeScale a few years earlier. This was one which our colleagues in the network team didn’t like, as they’d played around and made some network configuration changes, but the change went wrong and there was an outage for a very important R&D application. Using ThousandEyes, we could see immediately that they’d created a routing loop which they didn’t even realize themselves.”

Resolving Application-Layer Issues

Maarten also described how NXP had been able to leverage detailed Page Load tests to help drive performance improvement in key applications. “With ThousandEyes you can also have a very detailed breakdown of page loads. We were always having discussions with vendors for various applications who always said to us “your network is to blame” and advising us that we needed to fix our network.

With the one key application, which is very important for the R&D team we could use ThousandEyes to show the page loads for every detailed step and we could say to our supplier “look over here, you’re taking 3 seconds here, 4 seconds there, so your application is not right. Why does this take so long?” They didn’t have an answer so they had to go away and investigate. We identified 16 application faults in this application alone that the supplier had to go away and fix.”

Web application troubleshooting

Optimizing Skype for Business

One of the key areas of improvement NXP have seen has been on Microsoft Skype for Business, which is used for voice throughout NXP. Maarten explained, “Last January we had quite massive issues with Skype over in Austin, Texas, where we got a lot of complaints from users and management. So, we deployed an Enterprise Agent to Austin very quickly on a virtual machine. Within one hour we identified five issues we had between the site in Austin and the Skype back-end.”

In his presentation, Maarten goes on to focus on two particular areas where ThousandEyes helped to deliver extra insight. “In that hour, we found out that we had an ISP issue with Level 3. We route to the Microsoft network via the Internet from Austin and we saw that one router had a packet loss of 30%. We called Level 3, gave them the IP address and said this is wrong, please fix it. Level 3 called back and said they saw it too and it was fixed in one day and was fine after that.”

The second issue that Maarten focused on was around where traffic from NXP was entering the Microsoft network. “NXP was connected to Skype in Dublin, so everybody connected to the Skype back-end in Dublin. However, at that time Microsoft had been introducing multiple egress points and we started to experience Skype issues from Austin where 20%-30% of the calls were bad.

What we saw with ThousandEyes was that traffic from Austin was actually alternating between instances in Iowa, Texas and Canada! We contacted Microsoft and pointed this out and said we didn’t want multiple egress points and within one and a half weeks they’d fixed it.”

Latency issue with Skype egress

What are the Lessons Learned?

Maarten then shared what future plans NXP had for ThousandEyes, “Our next steps are to extend the test results to the business so that they can see the performance of their own site.” As a lot of people can add and run tests, he added “it became very popular and the agent can become overloaded, so it is very important to manage that”.

Maarten also commended the support from NXP had received from the ThousandEyes Customer Success team, that is available in the platform 24/7 via a chat function and provides support on interpret tests and all aspects of the platform; “We used the support organization heavily to help explain what we’re seeing. They were very helpful and knowledgeable, and they helped us a lot.”

If you would like to find out more about how ThousandEyes can be leveraged to monitor Office 365, why not watch one of our recent on-demand webinars – How to See and Resolve Office 365 Performance Challenges.

We’d also be happy to arrange a demo of the ThousandEyes platform, where we could walk you through some Office 365 and SaaS related use cases.

Subscribe to the
Network Intelligence Blog!
Subscribe
Back to ThousandEyes Blog