On this new episode of The New Stack Makers podcast, we spoke with Sean Armstrong, vice president of products at AppNeta, a SaaS company providing actionable, end-to-end network performance monitoring.
“The performance of your application,” Armstrong said, “is subject to how well the customer runs their own network and how well the internet between you and your customer is actually working. So you could get blamed for poor performance that’s completely outside of your control, and they use AppNeta to diagnose and troubleshoot those problems.”
With the explosion of SaaS companies and APIs providing building blocks for the modern IT stack, many if not most of its components are outside of their control. In addition, with the rise of remote working, it’s likely your users are no longer in a central office.
Traditional monitoring is about making sure that your devices are healthy. But you simply can’t do that when you don’t own the devices. “You need visibility after your firewall,” he said.
So all AppNeta clients start with deep packet inspection (DPI) analysis. What apps are on the network? What is on your network, what is competing for resources? What is changing about your network? Who is using the apps? Which apps are using resources?
Once all the pieces are discovered, they move to active testing, which they call 4D monitoring. This includes DPI analysis, net flow generation and packet capture along with active network testing and active web synthetic testing.
“So we know you’re having a problem with Office 365 by looking at the packet captures, so let’s actively test and figure out what’s causing that problem,” he said. It’s the delivery path and the load balanced cluster of servers that are changing second by second so just knowing what apps are on your servers is not enough.
The last piece, and most important, is providing root cause analysis and solution suggestions. For example, in the recent Facebook downtime, AppNeta actively monitored Facebook for its customers, some of whom connect with Facebook as part of their own products. The monitoring found that the network was healthy from all the transactions leading to Facebook, but the synthetic transactions found HTTP 500 errors in the app. So, he said, no major routing changes were suggested as the customers’ network was stable and the problem was clearly at the application layer.
Listen in to find out more about 4D monitoring, how machine learning is used, how much of the AppNeta stack is open source, and Armstrong’s advice for engineers.
In this Edition:
Feature image via Pixabay.