Modern distributed systems generate massive volumes of
telemetry data requiring both real-time processing and longterm analytical storage. Traditional single-database approaches
struggle to optimize for conflicting requirements: time-series
workloads demand high ingestion throughput and temporal
queries, while analytical workloads require flexible schema
support and complex aggregations [1].
This paper evaluates a dual-store telemetry backbone combining QuestDB [2] for time-series optimization and CrateDB [3] for structured analytics. Our approach addresses
the fundamental trade-off between ingestion performance and
query flexibility by routing telemetry streams to specialized
storage engines optimized for their respective workload characteristics.
A. Contributions
We make the following key contributions:
- Comprehensive benchmarking of QuestDB and CrateDB across ingestion, query, and cost dimensions using
realistic telemetry workloads - Dual-store architecture evaluation demonstrating performance characteristics and operational trade-offs
- Cost analysis including storage efficiency, retention policies, and operational overhead quantification
- Implementation guidelines for deploying dual-store
telemetry systems in production environments
B. System Architecture
Our dual-store telemetry backbone implements parallel ingestion to both QuestDB and CrateDB, as shown in Figure 1.
The ingestion pipeline routes telemetry streams via Influx Line
Protocol (ILP) to QuestDB for time-series optimization and
HTTP API to CrateDB for structured storage.