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 |
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?
