Skip to content

README.md

Latest commit

 

History

History
112 lines (84 loc) · 3.49 KB

README.md

File metadata and controls

112 lines (84 loc) · 3.49 KB

CPU and Memory Monitoring Scripts

This directory contains scripts for monitoring CPU and memory performance metrics on Linux systems using Intel Processor Counter Monitor (PCM).

Scripts

1. monitor-cpu.sh

Comprehensive CPU and memory monitoring script that tracks:

  • CPU utilization (usage percentage)
  • IPC (Instructions Per Cycle)
  • Core frequency (GHz)
  • L2/L3 cache miss counts
  • L2/L3 cache hit ratios
  • Local Memory Bandwidth (LMB) - MB/s
  • Remote Memory Bandwidth (RMB) - MB/s
  • I/O wait percentage

Usage

./monitor-cpu.sh [cores] [interval] [duration]

Arguments:
  cores    - Comma-separated list of cores (e.g., "0-7" or "0,1,2,3")
  interval - Sampling interval in seconds (default: 1)
  duration - Monitoring duration in seconds (default: 60)

examples

# Monitor cores 0-7 for 60 seconds (default interval: 1s)
./monitor-cpu.sh "0-7"

# Monitor specific cores with custom interval
./monitor-cpu.sh "0,1,2,3" 0.5 120

# Monitor multiple core ranges
./monitor-cpu.sh "0-7,16-23" 1 60

Output Format

========================================
CPU & Memory Monitor
========================================
Monitoring cores: 0,1,2,3,4,5,6,7
Sockets: 0
Monitoring duration: 60s
Collection interval: 1s
Output file: ./output/cpu-monitor-20260303-093921.log

Timestamp           | CPU |  UTIL% |   IPC | Freq(GHz) |     L2Miss |     L3Miss |   L2Hit% |   L3Hit% |  LMB(MB/s) |  RMB(MB/s) |    IOwait%
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
2026-03-03 09:40:21 |   0 |   5.12 |  0.89 |      3.62 |        120 |       9421 |    0.54  |    0.78  |        8.3 |        2.1 |       0.8%
2026-03-03 09:40:21 |   1 |   4.77 |  0.85 |      3.60 |        110 |       9010 |    0.55  |    0.77  |        7.9 |        2.0 |       0.8%
...

When run interactively, the terminal display is refreshed like `top` and always shows only the latest frame (one line per CPU).
The log file always appends every sampling frame (multi-line per frame: one line per CPU) for later analysis.

Output Columns

Column Description
Timestamp DateTime of sample
CPU Core number being monitored
UTIL% CPU utilization (0-100%)
IPC Instructions per cycle
Freq(GHz) Current core frequency
L2Miss L2 cache miss count
L3Miss L3 cache miss count
L2Hit% L2 cache hit ratio (0.00-1.00)
L3Hit% L3 cache hit ratio (0.00-1.00)
LMB(MB/s) Local memory bandwidth (MB/s)
RMB(MB/s) Remote memory bandwidth (MB/s)
IOwait% I/O wait percentage

2. monitor-cores-0-7-memory.sh

Simple memory bandwidth monitoring script specifically for cores 0-7.

Usage

./monitor-cores-0-7-memory.sh [interval] [duration]

Requirements

  • Intel PCM (Processor Counter Monitor) tools installed
  • Root privileges required (for accessing hardware performance counters)
  • Linux kernel with perf_event_paranoid set appropriately

Installation of PCM

PCM is typically pre-installed on systems with Intel Xeon processors:

# Check if PCM is available
/usr/local/sbin/pcm --help

Notes

  • Results are saved to /nvme4/xtang/scripts/cpu-monitor-<timestamp>.log
  • The socket mapping is automatically determined from lscpu -p
  • Memory bandwidth values represent bandwidth satisfied from local/remote memory controllers
  • Lower IPC or higher cache misses may indicate performance bottlenecks