System Development, Implementation, and Operations
Life cycle
Phases run planning/feasibility, requirements, design, build, test, implement, operate, and retire. Waterfall sequences them with formal signoffs; agile delivers iteratively and increases the risk of weak documentation and informal change control. Key control: segregation between development, test, and production - programmers should not have standing production access, and users, not developers, perform user acceptance testing against business requirements. Testing layers: unit (programmer), integration, system, regression, stress/load, and parallel.
Conversion strategies
Direct (big bang) is cheapest and riskiest; parallel is safest and most expensive because both systems run and outputs are reconciled; phased converts by module; pilot converts one location first. Questions often ask which strategy best mitigates the risk of undetected processing errors - parallel.
Change management and operations
Changes require documented authorization, testing, approval, and controlled migration to production, with emergency changes retroactively approved and reviewed. Operations covers job scheduling, monitoring, capacity, SLA management, incident versus problem management, and backup/restore. Full plus incremental backups (each captures changes since the last backup of any type) require the last full plus every incremental; full plus differential backups (each captures changes since the last full) require only the last full plus the latest differential. Backups must be stored offsite and restore-tested. End with a post-implementation review.
Worked example: A Sunday full backup plus daily differentials fails on Wednesday. Restore the Sunday full plus Tuesday's differential - not Monday's and Tuesday's, which is what an incremental scheme would require.
Trap: confusing who tests what, and assuming a tested backup equals a recoverable backup.