Integrated Capstone — High-Volume Service Remediation
Situation
A company launches a high-volume service with:
- one Production database;
- a scaled-down Test environment;
- daily backups stored on the same filesystem as the database;
- no restore tests;
- month-end load spikes;
- customer Production data copied into sandboxes;
- a plan to keep every record forever.
Leadership asks for a Chapter 6 remediation sequence.
Strong sequence
- Classify business criticality and define service/continuity expectations and restoration priorities.
- Separate backup storage, set SLA-backed frequency, secure appropriate off-site copies, and test recovery.
- Make Test sufficiently Production-like for the tests being performed; do not use Production as the test environment.
- Review Sandbox isolation and protected-data controls; mask/restrict Production-derived customer data or use synthetic/purpose-built data.
- Model capacity and growth including indexes, logs, mirrors/redundancy, and the storage impact of retention.
- Analyze usage patterns and performance bottlenecks; tune from evidence before structural redesign/denormalization.
- Define deliberate retention → archive → purge policies rather than indefinite retention by default.
- Establish monitoring/metrics, configuration control, and scripted change promotion.
Why this order matters
The weakest response is a technology-only redesign. Chapter 6 treats Data Storage and Operations as a controlled service: business continuity and requirements drive technology, environments, backup/recovery, capacity, performance, lifecycle, data protection, and governance.
Changed-fact test
For each step, state one changed fact that would alter the response. Examples: - off-site restores already tested successfully → focus shifts to another continuity weakness; - Test is already Production-equivalent → focus shifts from infrastructure similarity to test scope/data; - records have indefinite legal-retention requirement → “purge after N years” is no longer appropriate; - slowdown correlates to stale statistics after volatility → optimizer/statistics triage becomes immediate.
Source: Chapter 6, pp. 166–208.