Performance must be measured, not implied.
RayoDB evaluations compare latency, throughput and footprint against the system already running the same workload.
Where the public comparison stands.
The broad comparison program is still in progress. Results will appear when the same evidence supports both performance and workload coverage.
Until then, each evaluation runs side by side with systems such as ClickHouse, using the same queries, data and service objective.
Four dimensions, measured together.
Response time
Median and tail latency for individual queries, with cache state and timing scope declared.
Concurrent capacity
Throughput and response times as more users, dashboards and scheduled jobs share the system.
Infrastructure
Servers, processors, GPUs, memory, storage and power required to sustain the measured service level.
Correctness
Identical data, equivalent SQL and verified answers. A fast wrong result does not count.
A side-by-side evaluation in three steps.
The comparison is designed around the service your business needs, not an isolated best-case query.
Replay the real workload.
Load representative data and run the queries, joins and update patterns that matter in production.
Add realistic concurrency.
Increase clients while recording throughput, median latency, tail latency and every error.
Compare the footprint.
Measure the hardware required by each system to maintain the same response-time objective.
Compare it with your current stack.
Early-access partners run a controlled evaluation with RayoDB engineers and keep the complete result.