How GrafanaCloud Is Enabling Packet’s Teams, from Engineering to Sales
“The first priority when we founded Packet was getting our platform functional,” says Nathan Goulding, Packet’s SVP of Engineering. “Logging and monitoring weren’t as robust as we needed them to be, and we didn’t have the necessary metrics set up out of the gate. As we started to scale the platform, we felt the pain that comes from not having the visibility we needed, and realized that this was a massive blind spot for us. So we had to ask ourselves, ‘What’s the approach that we should take?’”
Single Pane of Glass
Packet—which deploys its application via container orchestration—had begun collecting data from Observium (to monitor switch ports and traffic), Elasticsearch, and Prometheus (for container status, host metrics, and container metrics). But there was no generic dashboard that could hook up to the different data sources. To solve that problem, the team looked at several commercial products, but “most of these products were not offering the metrics and TSDB-based data stores and dashboarding, which is really where our pain point was,” says Goulding.
Focus on Apps, not Ops
The team was already familiar with Grafana as a software package, and the ideal solution presented itself when GrafanaLabs launched its GrafanaCloud product in 2017. “GrafanaCloud was very compelling to us because of our company’s size and needing to focus our limited resources on building out our core product,” says Goulding. “Setting up the necessary infrastructure for TSDB and metrics is feasible, but in practice it’s challenging to maintain and optimize and monitor and scale it successfully. A lot of the things that Grafana has done under the hood to optimize and in some cases even rewrite the internal workings has made it even more cost effective for us, and something that we simply would not have had the time to do.”
“Not having to manage this ourselves,” he adds, “quickly moved us from the evaluation phase to using it extensively for production metrics.”
"Not having to manage this ourselves, quickly moved us from the evaluation phase to using it extensively for production metrics."
Dashboards for Every Team
At the company’s lower Manhattan office, Grafana dashboards are prominently displayed on TV monitors mounted around the “bullpen,” and a number of staffers have dashboards up at their desktops.
Packet has built around 40 different dashboards for its teams to use as needed. “We have the platform team that looks at all of the host metrics and our netops team that looks at traffic,” says Goulding. “We have the engineering team that looks at API metrics, and we have the finance team and business sales folks that look at things like inventory and capacity and real-time public cloud usage data. There’s a good chance that most people at Packet will be looking at a Grafana dashboard at some point during their work day.”
The most valuable metrics that are visualized on Packet’s Grafana dashboards, Goulding points out, are fairly standard: the number of API queries per second, the top API methods, the top users, the top IP addresses, the slowest queries. But “being able to correlate container utilization, disk space, CPU, API, platform metrics, and application metrics all together in a single place is what’s most crucial,” he says. “There’s a lot of insight and context that you can gain when you’re able to correlate these things quickly.”
That desire to integrate all of the different data sources had been a primary driver in Packet’s search for a hosted metrics solution. And that is just what Grafana has enabled Packet to do. “Being able to, in a single dashboard, pull in data from a variety of different data sources,” says Goulding, “provides the biggest value for us.”