The Redwood City, California-based company develops software that enables business analysts to prepare their data to be used for analytics. Paxata’s environment runs 70 to 80 hosts at any given time and consists of 25 production servers. It also recently began moving to a containerized environment with Kubernetes and Dockers. Now, it uses them extensively for test-dev and quality assurance and is migrating to use containers for production as well.
While Paxata had been using monitoring tools, they were too traditional for what the company needed. It was using Sysdig cloud-based APM, which primarily focused on infrastructure metrics but didn’t help the company get to the root of its application problems.
According to David Levinger, VP of development and cloud operations at Paxata, these application problems were extremely time-consuming and tedious to get to the bottom of. He reflected that while the company could tell whether or not code was running or not running, or whether it’s system was up or down, it could not tell how it was performing — and when it wasn’t performing, they didn’t know why or where. The company needed deeper insight.
“One of the key differentiators between us and our key competitor is interactivity at scale… and that interactivity peice is really where Instana came in for us,” said Levinger. When deploying Instana, the company was “trying to figure out how do we provide a better service, how do we make that interactivity smoother, more responsive, and kind of dig into our code.”
Levinger said that there was an “immediacy of value” that Paxata got from Instana.
Within a day or two, Paxata had a massive set of performance problems on the code-level that it hadn’t realized were there. “All of the sudden we were seeing it — we could see exactly where it was misbehaving and when it was misbehaving,” he said.
Shortly after its initial deployment, Paxata also deployed Instana’s software development kit (SDK) to gain insight at its websocket layer. Levinger said 15 to 20 major performance issues were immediately exposed to its developer team. Not only that, but its developers were able to see how and why the problems were manifesting and fix them at the code-level.
Levinger said the greatest benefits, aside from the problems it initially detected, were Instana’s ability to perform code-level traces and to alert problems before they’re detected by customers. He said that Paxata set up a production dashboard, segregated it based on users and tenants, and was able to add traces through the filtered lenses to identify issues.
Paxata did evaluate other APM vendors, namely AppDynamics and Stackify. However, it settled on Instana because of its support for containerization, websocket, and Apache Spark — all technologies that it relies on.
On the container level, Levinger said that the ability to deploy the monitoring tool natively and “hook-in at the Kubernetes level, in the right way, in the way that Kubernetes works, was very valuable for us.”
Levinger’s only regret? That he didn’t deploy the tool sooner. “I would say one thing that Instana showed us is you don’t know what you don’t know,” he said, adding that the sooner the instrumentation is in place and the sooner it’s built into your software development life cycle, the more virtuous it becomes.