The Six-Minute Boot: Triage in a 40-Seat Computer Lab
A college computer laboratory has 40 desktop machines, each with 4 GB RAM, a 5400 rpm hard disk and Windows 10. The lab opens at 08:00 and the first class arrives at 08:05. Students report that machines take 6-9 minutes to reach the desktop and that application windows freeze for 20-30 seconds when switching. Each machine has 12 programs set to start at login, a scheduled antivirus full scan at 08:00, and disks that are 88-94% full. Nothing has been reimaged in three years. Two technicians are available, and the department has KES 150,000 for the whole term. Eight machines now show SMART reallocated-sector warnings and shut down randomly.
1. Rank the likely causes of the slow startup and justify the ranking using the evidence given, not generic lists.
Read the worked answer
Ranked diagnosis
- Disk I/O saturation from the 08:00 antivirus full scan. This is the strongest candidate because the symptom is time-bound: it begins exactly when the lab opens and coincides with the scan. A 5400 rpm disk delivers roughly 80-120 IOPS; a full scan plus 12 startup programs plus profile load will queue requests, producing exactly the 20-30 second freezes described.
- Near-full system disks (88-94%). Windows needs free space for the page file, hibernation file and temporary files. At above ~85% full, defragmentation and update staging degrade sharply, and the page file cannot grow when RAM (4 GB) is exhausted by 12 startup programs.
- RAM pressure. 4 GB with a dozen login-time programs and a browser will page to disk, compounding item 1.
- Three-year-old image. Accumulated updates, orphaned registry entries and obsolete drivers explain the baseline slowness that exists even outside 08:00.
Deliberately ranked lower: hardware age alone. Age is not a mechanism; the SMART warnings explain crashes on 8 machines, not the 30-second freezes on all 40. Do not let one visible symptom (crashes) absorb the diagnosis of a different symptom (latency).
2. Propose a two-week remediation sequence within KES 150,000, and state at least two actions you would deliberately defer or reject with reasons.
Read the worked answer
Week 1 (low cost, high impact)
- Move the antivirus full scan to 18:00 and switch to quick scans daily with a full scan weekly: KES 0, removes the single biggest I/O competitor at 08:00.
- Trim login startup items from 12 to 4 on all 40 machines via policy: KES 0.
- Clear temp files, move student data to a network share, target below 75% disk usage: KES 0 in labour beyond technician time.
- Replace the 8 SMART-flagged disks or, better, reimage those 8 onto new SSDs.
Budget check. 8 x 480 GB SSD at ~KES 5,500 = KES 44,000; 8 x 4 GB RAM upgrade at ~KES 3,000 = KES 24,000; cloning dock KES 12,000; contingency KES 20,000. Total ≈ KES 100,000, leaving KES 50,000.
Week 2 - Build one reference image, validate on 4 machines, then roll out to 32 in batches of 8 per evening. Use free space to lift RAM on the 8 busiest machines if the pilot shows paging is still the binding constraint.
Deferred or rejected
- Buying 40 SSDs (≈KES 220,000): breaches the budget and ignores that the 08:00 scan is free to fix.
- Full lab reimage in one night: with two technicians, a failed image leaves the lab unusable for a morning class. Batch it.
- Registry cleaners: no measurable benefit and they risk breaking the image.
3. For the 8 machines with SMART warnings, would you repair or replace, and what would change your decision?
Read the worked answer
Recommendation: replace, do not repair.
Reallocated sectors indicate the platter surface is already failing; this is a wear-out failure mode, not a configuration error. A repair (reinstall, chkdsk, defrag) cannot restore a degrading surface, and the department would pay technician hours twice: once now, once when the disk dies mid-lesson.
Cost logic. Repair costs roughly KES 0 in parts but 2-3 technician hours per machine, and leaves an expected failure with no warning. Replacement at ~KES 5,500 for an SSD costs less than the labour it displaces and also fixes the latency problem on those 8 seats.
What would change the decision
- If the SMART count is stable across two readings a week apart and the only flag is a non-critical attribute, the disk may be a monitoring case rather than an immediate replacement.
- If the machines are due for full replacement within one term under a capital plan, patch minimally and push the spend into that plan.
- If the budget were already committed, I would consolidate the 8 failing units into one room so failures are contained and lessons are not disrupted across the whole lab.