Two weeks ago Axiom headed to Portland, Oregon as a sponsor of Monitorama — a popular conference for monitoring and observability practitioners now in its 10th year.
Axiom CTO Seif Lotfy presents an introduction to Axiom at Monitorama 2023 in Portland.
I soaked up learnings from attendees, vendors, and all the talks. Here are my top five takeaways:
1. Strong observability practices depend on connected, multi-modal events. OTel and eBPF are winners.
Logs, metrics and traces all help teams follow the breadcrumbs to answer why a problem is happening. So there was universal love for portable, standarized formats like OpenTelemetry that make event correlation easier, and growing interest in eBPF sensors that simplify instrumentation. Teams want more events, and more connected events. All in the name of actionable, directed search.
2. Economic climate and cost sensitivity around monitoring drives innovation.
At Enterprise scale, the tsunami-wave of observability data means the cost of monitoring is hurting; it can now be as high as 20-30% of cloud spend. Some are conceding cardinality, or dropping events on the floor by filtering and sampling through pipelines. Others, like Netflix, are fighting to balance cost, scale and latency head-on, leaning on tools like ClickHouse in cloud-native architectures that reflect the very innovation behind Axiom. Huge savings await.
3. SLO-based alerts cut through the noise and promote action.
We heard just how much solving fatigue from false positives and redundant alerts matters. One answer is SLO-based alerts that are biased towards action through a connection to our observability data, and run-books that guide teams to remediation. Automation can ease alert overload, keeping teams focused on the notifications that count. No shock collars needed (IYKYK).
4. AI gets dunked on. But has its place.
As the rest of tech fawns AI, speakers couldn’t resist poking some fun. There are two kinds of AIOps companies, we were told: “dead ones and new ones”, with AI being little more than “three if statements in a trench coat”. The real takeaway, though, was that if AI and LLMs can help us explore unfamiliar systems and get to insight faster, they increase leverage, productivity and value. And that’s a trench coat we’d all welcome.
5. The math behind monitors matters. Teachers rejoice.
Log-based histograms. Cumulative distribution functions. The differences between arithmetic, geometric, and harmonic means. It all featured. The lesson was that we can solve problems more effectively when we appreciate the application of relevant math and statistics. And as tool-makers, we have work to do to make this all more accessible. Those textbooks mattered, after all.
This is a space full of passionate practitioners and venturous vendors.
I’m grateful to have made new friends at companies like groundcover, Armory, Honeycomb and Shoreline, and excited to apply these learnings as we continue solving important problems at Axiom. Onwards!