Observability - a year in review

In 2025, the observability landscape saw significant evolution, driven by increasing data volume and cardinality, the rise of tracing as a primary signal, and the consolidation of OpenTelemetry as a standard. ClickStack, introduced by ClickHouse, aimed to make high-performance observability more accessible, leveraging HyperDX UI to optimize query execution for logs, traces, and metrics. Volume alone was no longer the main challenge; instead, high cardinality emerged as a key constraint, especially with AI workloads introducing complex telemetry. Tracing became essential for understanding complex systems, often overshadowing traditional logs and metrics. OpenTelemetry became the default for new services while older systems were gradually retrofitted, despite challenges in adoption due to complex configurations and evolving standards. Open source observability solutions gained traction as teams migrated from existing platforms, valuing cost predictability and data ownership over polished user experiences. AI-driven Site Reliability Engineering (SRE) tools showed potential but remained underdeveloped, with human judgment still crucial. Data quality became a pressing issue as teams focused on the utility and cost of telemetry signals, marking a shift from mere data collection to meaningful analysis.