Monitoring Fundamentals — Trivia & Interesting Facts¶
Surprising, historical, and little-known facts about monitoring fundamentals.
Nagios was created in 1999 and is still running in production at thousands of organizations¶
Ethan Galstad created NetSaint in 1999 (renamed Nagios in 2002 due to trademark issues) as a host and service monitoring tool for Linux. Despite being considered legacy technology, Nagios and its forks (Icinga, Naemon, Shinken) still run in production at thousands of organizations worldwide. The plugin-based architecture, with over 5,000 community plugins, proved remarkably durable.
The "four golden signals" originated at Google and simplified monitoring for everyone¶
Google's SRE book (2016) defined the Four Golden Signals: latency, traffic, errors, and saturation. Before this framework, monitoring was ad-hoc — teams monitored whatever metrics were easy to collect, often focusing on CPU and memory while ignoring user-facing indicators. The four golden signals provided a simple, universal framework that transformed how the industry approaches monitoring.
Prometheus was inspired by a Google-internal monitoring system called Borgmon¶
Matt Proud and Julius Volz created Prometheus at SoundCloud in 2012, directly inspired by Google's Borgmon monitoring system that they had used as former Google employees. Prometheus adopted Borgmon's dimensional data model (labels), pull-based collection, and powerful query language. It became a CNCF graduated project in 2018 and the de facto standard for Kubernetes monitoring.
SNMP has been monitoring network devices since 1988¶
Simple Network Management Protocol (SNMP) was first published as RFC 1067 in 1988. Version 1 transmitted community strings (passwords) in plaintext. Despite known security issues, SNMPv2c (also with plaintext community strings) remains the most widely deployed version. SNMPv3, which added encryption and authentication, was standardized in 1998 but adoption has been slow due to configuration complexity.
The USE method was created because engineers were monitoring the wrong things¶
Brendan Gregg created the USE (Utilization, Saturation, Errors) method in 2012 after observing that engineers frequently missed performance problems because they were checking the wrong metrics. The USE method provides a checklist: for every resource (CPU, memory, disk, network), check utilization, saturation, and errors. This systematic approach catches problems that ad-hoc monitoring misses.
Time-series databases exploded from 2 options to over 30 in a decade¶
In 2010, the primary options for storing monitoring data were RRDtool (1999) and Graphite's Whisper (2008). By 2020, there were over 30 time-series databases including InfluxDB, TimescaleDB, VictoriaMetrics, Thanos, Cortex, M3, QuestDB, and TDengine. This explosion was driven by the monitoring data growth that accompanied the microservices and cloud-native movements.
Cardinality explosions have brought down monitoring systems at major companies¶
High-cardinality labels (like user IDs or request IDs attached to metrics) can cause the number of unique time series to explode exponentially. Prometheus instances have been crashed by a single misconfigured metric with a high-cardinality label generating millions of time series. Datadog, New Relic, and other vendors now explicitly charge based on custom metric cardinality because of the computational cost.
Monitoring was originally done by humans watching consoles 24/7¶
Before automated monitoring, early computing centers employed operators who literally watched blinking lights and console outputs around the clock. IBM mainframe operators monitored system health through console messages on physical terminals. The transition from human observation to automated monitoring happened gradually through the 1980s and 1990s, and the "NOC" (Network Operations Center) with wall displays is a direct descendant of this tradition.
Grafana started as a Kibana fork in 2014 and became a $6 billion company¶
Torkel Odegaard forked Kibana 3 in 2014 because he wanted a better time-series visualization tool for Graphite data. He stripped out the log-specific features and focused on dashboarding. Grafana Labs, the company built around the project, reached a $6 billion valuation by 2022. Grafana's plugin system supports over 150 data sources, making it the most popular open-source monitoring dashboard.