Metrics

The following list contains the core metrics reported by Linaro MAP.

Only available metrics are exported to JSON. For example, if there is no Lustre filesystem then the Lustre metrics will not be included. If any custom metrics are loaded, they will be included in the JSON, but are not documented here.

For more information on the metrics, see Metrics view

  • CPU Instructions: see CPU instructions

    • instr_fp

    • instr_int

    • instr_mem

    • instr_vector_fp

    • instr_vector_int

    • instr_branch

    • instr_scalar_fp: The percentage of time each rank spends in standard x87 floating-point operations.

    • instr_scalar_int: The percentage of time each rank spends in standard integer operations.

    • instr_implicit_mem: Implicit memory accesses. The percentage of time spent executing instructions with implicit memory accesses.

    • instr_other: The percentage of time each rank spends in instructions which cannot be categorized as any of the ones given above.

  • CPU Time: see CPU time

    • cpu_time_percentage: See CPU time in CPU time

    • user_time_percentage: See User-mode CPU time in CPU time

    • system_time_percentage: See Kernel-mode CPU time in CPU time

    • voluntary_context_switches: See Voluntary context switches (1/s) in CPU time

    • involuntary_context_switches: See Involuntary context switches (1/s) in CPU time

    • loadavg: See System load in CPU time

  • I/O: see I/O

    • rchar_rate: See POSIX I/O read rate (B/s) in I/O

    • wchar_rate: See POSIX I/O write rate (B/s) in I/O

    • bytes_read: See Disk read transfer in I/O

    • bytes_written: See Disk write transfer in I/O

    • syscr: See POSIX read syscall rate in I/O

    • syscw: See POSIX write syscall rate in I/O

  • Lustre

    • lustre_bytes_read: Lustre read transfer (B/s)

    • lustre_bytes_written: Lustre write transfer (B/s)

    • lustre_rchar_total: Lustre bytes read

    • lustre_wchar_total: Lustre bytes written

  • Memory: see Memory

    • rss: See Memory usage in bytes (Resident Set Size) in Memory

    • node_mem_percent: See Node memory usage in Memory

  • MPI: see MPI calls

    • mpi_call_time: See MPI call duration (ns) in MPI calls

    • mpi_sent: See MPI sent in MPI calls

    • mpi_recv: See MPI received in MPI calls

    • mpi_calls: Number of MPI calls per second per process

    • mpi_p2p: See MPI point-to-point in MPI calls

    • mpi_collect: See MPI collective operations in MPI calls

    • mpi_p2p_bytes: See MPI point-to-point bytes in MPI calls

    • mpi_collect_bytes: See MPI collective bytes in MPI calls

  • Accelerator (NVIDIA): see Accelerator

    • nvidia_gpu_usage: See GPU utilization in Accelerator

    • nvidia_memory_used_percent: See GPU memory usage in Accelerator

    • nvidia_memory_used: GPU memory usage in bytes

  • Accelerator (AMD): see Accelerator

    • rocm.gpu_utilization: See GPU utilization in Accelerator

    • rocm.gpu_memory_used_percent: See GPU memory usage in Accelerator

    • rocm.memory_utilization: See GPU memory utilization in Accelerator

    • rocm.gpu_memory_used: GPU memory usage in bytes

  • Energy: see Energy

    • nvidia_power: See NVIDIA GPU power usage (mW/node) in Energy

    • rocm.gpu_power: See AMD GPU power usage (mW/node) in Energy

    • rapl_power: See CPU power usage (W/node) in Energy

    • system_power: See System power usage (W/node) in Energy

    • rapl_energy: CPU energy, integral of rapl_power (J)

    • system_energy: CPU energy, integral of system_power (J)