When an index changes the plan but not the outcome
Why plan improvement and latency improvement are different claims—and what to do when the first remedy fails.
Working notes
Reliability without the theater
Practical arguments, tested where possible.
Writing about database reliability should expose assumptions, evidence, trade-offs, and limitations. These are the first subjects in the Cloud DBRE Tech editorial pipeline.
Local qualification is not production approval
A practical boundary between useful experiments and decisions that can survive real workload, ownership, and failure modes.
Read models are reliability decisions, not just performance tricks
The hidden refresh, staleness, recovery, ownership, and cost questions behind a fast reporting path.
What a safe database assistant should be allowed to do
A capability model for evidence collection, diagnosis drafts, change proposals, review gates, and prohibited production actions.
From incident artifacts to reusable diagnostic evidence
How logs, plans, metrics, hypotheses, rejected candidates, and verification results can shorten the next investigation.
The database reliability review I want before a migration
A concise examination of workload isolation, recovery objectives, observability, cost, change control, and operational ownership.
Source discipline
External SRE, database, and reliability sources will be used to support specific claims—not to manufacture volume. Each article should add an applied argument, a concrete example, a useful synthesis, or a clearly labeled opinion. Primary documentation and original research take precedence over recycled summaries.