3.5.2. Storage Monitoring and Optimization
💡 First Principle: Storage problems announce themselves before they become critical—if you're watching. Slow I/O, rising queue depth, and capacity approaching 80% are all warning signals. Monitoring storage proactively converts potential outages into scheduled maintenance.
Key Storage Metrics to Monitor
| Metric | What It Indicates | Action Threshold |
|---|---|---|
| Capacity utilization | How full the volume is | Alert at 70%, plan at 80%, critical at 90% |
| IOPS (Input/Output Operations per Second) | Storage workload intensity | Compare against baseline; queue depth growth indicates saturation |
| Throughput (MB/s) | Data transfer rate | Near-maximum for the drive/RAID type indicates bottleneck |
| Latency | Time per I/O operation | HDD: alert >20ms; SSD: investigate >1ms, alert >5ms (normal is well under 1ms) |
| RAID array health | Drive/array status | Any degraded state requires immediate attention |
Growth rate beats static thresholds. The percentages above assume ordinary growth. Divide the remaining headroom by the monthly growth rate to get the months left before critical, and start the expansion while that runway is still comfortably longer than the procurement lead time—under fast growth that point can come well before the 80% planning mark.
Deduplication and Compression Monitoring
Both features reduce usable storage at a CPU cost. Monitor:
- Deduplication ratio (actual savings achieved)
- CPU overhead during deduplication processing windows
- Whether compression/deduplication jobs complete before the next backup window
⚠️ Exam Trap: Deduplication is typically a post-process (runs after data is written, usually during off-hours). Inline deduplication (real-time during writes) has much higher CPU impact. The exam may present a scenario where a server is slow during backup windows—deduplication competing with backup is a common cause.
Reflection Question: A file server's disk utilization has been consistently growing at 5% per month for the past six months. Current utilization is 72%. What action should you take, and by when?
