Attendance System Cutover Under Time Pressure
Riverside Academy has 1,850 students and 140 staff. Its attendance system runs on a Windows Server 2008 R2 machine with SQL Server 2008; vendor support ended two years ago. The board has bought a cloud attendance service and wants a cutover before the new term in six weeks. Seven years of records - about 2.3 million rows - must move. The school network is 100 Mbps, but the server room switch is 10/100 and one uplink runs at 35% utilisation at peak. A trial export lost 47 duplicate rows and 12 rows with invalid dates. Forty percent of staff rate their IT confidence as beginner. The head teacher insists that no attendance session may be lost during transition, and the vendor offers no automatic rollback.
1. Recommend a changeover strategy for the attendance system. Justify your choice by comparing it with at least one alternative, and state one condition that would make you change your recommendation.
Read the worked answer
A phased changeover by year group is the safest fit. Run the cloud system and the legacy system in parallel for two weeks, starting with Year 7 and one sixth-form group, then add year groups over three weekends. This limits the blast radius: if the export has bad dates or duplicates, only a small group is affected, and staff can still record attendance in the legacy system. A direct cutover is cheaper and faster, but the trial export already lost 47 duplicate rows and 12 invalid dates, and the vendor offers no automatic rollback. A full parallel run for all 1,850 students would avoid data loss but would double data entry and create reconciliation conflicts across 140 staff, many of whom are beginners. The condition that would change the recommendation is evidence that the vendor can provide a tested rollback and a reconciled migration report; in that case a direct cutover during a holiday, with a two-day pilot, becomes defensible.
2. Identify two risks in this migration and explain a specific mitigation for each. Your answer should distinguish between data risks and human risks.
Read the worked answer
Data risk: duplicate or invalid records corrupt attendance history. The trial export lost 47 duplicate rows and 12 invalid dates, which suggests weak constraints in the legacy schema. Mitigation: run a validation script before migration - presence check on student ID, range check on date, consistency check that attendance date falls within the academic term, and a duplicate rule on student ID plus date plus session. Then verify by comparing row counts per term and by checksumming a sample of 200 records against manually checked legacy pages. Human risk: 40% of staff are beginners and may mistype or avoid the new system. Mitigation: role-based training in the two weeks before go-live, a one-page quick reference, and two “floor walkers” on the first three mornings. Also keep a read-only legacy view for four weeks so staff can check past records without re-entering data.
3. The head teacher asks whether the 100 Mbps network is sufficient. Explain what you would measure and how the 10/100 switch and 35% uplink utilisation affect the answer.
Read the worked answer
Bandwidth alone does not decide this. I would measure concurrent users at the peak attendance period, average attendance-record size, and the cloud service’s request pattern. If 140 staff each submit a small form, the payload may be only a few hundred kilobytes per minute, which is trivial on 100 Mbps. But the 10/100 switch caps any single wired device at 100 Mbps and, more importantly, may be a shared half-duplex or congested segment. The 35% uplink utilisation at peak leaves headroom now, but a cloud cutover adds synchronisation traffic, and a 2.3 million-row initial migration is a burst, not a steady load. I would schedule the bulk migration outside teaching hours, use a wired connection for the migration host, and monitor switch port utilisation and error counters. If the switch shows high collision or discard rates, replace it before go-live; otherwise the 100 Mbps link is probably sufficient for daily attendance.