Ontological Groupings for sysrec Fields
Created by: DrQz
- Rename some fields, e.g., Processor or Cores (instead of CPU) - may want it short
- Current CPU field really contains O/S scheduler metrics, e.g., nice pct
- Make MEM fields a separate KDR record, e.g., mermen, while retaining essentials in 'sysrec'
- Many metrics (especially for passive resources) are zeros
- Are all the O/S metrics in sysrec really necessary?
Current KDR metrics organization:
FIELDS:
CPU
#01 timestamp : seconds since Epoch, time
#02 cpupct : cpu utilization, across all cpus, number
#03 sumpct : sum of all cpus utilization, number
#04 headpct : headroom cpu available, all cpus, number
#05 userpct : cpu utilization, user space in percent, number
#06 nicepct : cpu utilization, user space with nice priority, number
#07 sysct : cpu utilization, system space, number
#08 idlepct : cpu utilization, idle state, number
#09 iowaitcpt : cpu percentage in idle state because an I/O operation
is waiting to complete, number
#10 irqpct : cpu percentage servicing interrupts, number
#11 softirqpct : cpu percentage servicing softirqs, number
#12 stealpct : cpu percentage of time spent in other operating systems
when running in a virtualized environment, number
#13 runqsz : run queue length, tasks waiting for run time, number
#14 plistsz : tasks in the task list, number
A better ontological organization for "CPU" would be:
FIELDS:
SAMPLE
#01 timestamp : seconds since Epoch, time
CORES
#02 cpupct : cpu utilization, across all cpus, number <--- %cpu AVERAGED over all cores
#03 sumpct : sum of all cpus utilization, number <--- TOTAL %cpu can be > 100%
#04 headpct : headroom cpu available, all cpus, number
#05 userpct : cpu utilization, user space in percent, number <--- s.b. user TIME
#07 sysct : cpu utilization, system space, number <--- s.b. syspct? and sys TIME
#08 idlepct : cpu utilization, idle state, number <--- s.b. idle TIME
#09 iowaitcpt : cpu percentage in idle state because an I/O operation
is waiting to complete, number
#12 stealpct : cpu percentage of time spent in other operating systems
when running in a virtualized environment, number
SCHED
#06 nicepct : cpu utilization, user space with nice priority, number
#10 irqpct : cpu percentage servicing interrupts, number
#11 softirqpct : cpu percentage servicing softirqs, number
#13 runqsz : run queue length, tasks waiting for run time, number
#14 plistsz : tasks in the task list, number
#44 avg1 : LA of the last minute, number
#45 avg5 : LA of the last 5 minutes, number
#46 avg15 : LA of the last 15 minutes, number
Edited by Stefan Parvu