In AIX, performance monitoring is focused on identifying CPU bottlenecks, memory pressure, disk I/O issues, and network congestion—especially in enterprise workloads running on systems like IBM Power Systems.

Here are the key performance metrics you should monitor:

🧠 1. CPU Utilization & Saturation

Key indicators:

• %usr → user process CPU usage
• %sys → system/kernel CPU usage
• %idle → idle CPU time
• %iowait → CPU waiting for I/O

👉 What it tells you:

• High %usr/%sys → CPU-heavy workload
• High %iowait → storage bottleneck
• Low %idle → CPU saturation

💾 2. Memory Usage & Paging Activity

Key metrics:

• Free memory vs used memory
• Paging space utilization
• Page-in / page-out rates

👉 What it tells you:

• High paging = memory pressure
• Excess page-outs = performance degradation
• Stable memory = healthy workload behavior

📦 3. Disk I/O Performance

Key metrics:

• Disk read/write IOPS
• Average service time
• Queue depth
• Disk utilization %

👉 What it tells you:

• High service time → slow storage
• High queue depth → I/O congestion
• Consistent I/O → stable system performance

🌐 4. Network Performance

Key metrics:

• Packets per second (PPS)
• Bandwidth utilization
• Packet errors/drops
• TCP retransmissions

👉 What it tells you:

• Drops/retransmits → network congestion or misconfiguration
• High utilization → potential bandwidth limits

⚙️ 5. Process & Workload Metrics

Key metrics:

• Top CPU-consuming processes
• Process wait states
• Thread counts
• Run queue length

👉 What it tells you:

• Which applications are driving system load
• Whether workloads are over-subscribed

🧩 6. Virtualization Performance (LPAR-Level)

In Power environments using IBM PowerVM:

• Entitled capacity vs used capacity
• CPU steal time (shared processor mode)
• LPAR dispatch wait time

👉 What it tells you:

• Whether virtual machines are properly provisioned
• Whether resource contention exists

🔄 7. I/O Wait & System Contention

• %iowait (CPU waiting for I/O)
• Lock contention
• Run queue delays

👉 What it tells you:

• Whether performance is blocked by disk or synchronization issues

🧠 8. System Load Average

• 1-minute, 5-minute, 15-minute load averages

👉 What it tells you:

• Overall system demand vs capacity
• Whether workload is stable or spiky

📊 9. Application-Level Performance

• Transaction response time
• Throughput (requests/sec)
• Error rates

👉 What it tells you:

• End-user experience
• Whether system performance meets SLA requirements

🔐 10. Resource Contention & Limits

• CPU entitlement limits
• Memory caps (in LPARs)
• File descriptor usage

👉 What it tells you:

• Whether system limits are restricting performance

🔁 11. High Availability & Failover Metrics

In systems using IBM PowerHA:

• Failover time
• Cluster health status
• Node heartbeat status

👉 What it tells you:

• System resilience and recovery readiness

📌 Common AIX Monitoring Tools

• topas → real-time system view
• nmon → detailed performance tracking
• vmstat → CPU + memory stats
• iostat → disk I/O performance
• netstat → network activity

🔍 Bottom Line

In AIX environments, the most critical performance metrics are:

• CPU utilization and iowait
• Memory usage and paging activity
• Disk I/O latency and throughput
• Network packet and bandwidth performance
• Virtualization (LPAR) resource usage
• Application response times and load