All Along the Watchtower: Achieving the Trinity of Observability in Cloud with DiTing

Observability is crucial for diagnosing and troubleshooting cloud systems as they grow in scale and complexity. However, telemetry data are often stored and processed in siloed systems, leading to high latency, redundant data movement, and low operational efficiency. To address these challenges, we present DiTing, an observability framework that unifies the storage and processing of heterogeneous telemetry data, including logs, metrics, and traces. The key idea of DiTing is to harvest underutilized cloud resources for cost-effective processing while relying on centralized storage systems for reliable data persistence and failover. We have deployed DiTing in production at scale, and DiTing can achieve sub-second data ingestion, high-throughput query processing, and up to 65× lower CapEx than existing solutions.