stats – display graphs of system activity


stats [ -option ] [ machine ... ]


Stats displays a rolling graph of various statistics collected by the operating system and updated once per second. The statistics may be from a remote machine or multiple machines, whose graphs will appear in adjacent columns. The columns are labeled by the machine names and the number of processors on the machine if it is a multiprocessor.

The right mouse button presents a menu to enable and disable the display of various statistics; by default, stats begins by showing the load average on the executing machine.

The lower-case options choose the initial set to display:

b battery

percentage battery life remaining.  

c context

number of process context switches per second.  

d draw

draw memory allocation size in bytes.  

e ether

total number of packets sent and received per second.  

E etherin,out

number of packets sent and received per second, displayed as separate graphs.  

f fault

number of page faults per second.  

i intr

number of interrupts per second.  

I idle

system load, % time in idle, and % time in interrupts. The last two are averaged over all processors on a multiprocessor.  

k kern

kernel memory allocation size in bytes.  

l load

(default) system load average. The load is computed as a running average of the number of processes ready to run, multiplied by 1000.  

m mem

total pages of active memory. The graph displays the fraction of the machine’s total memory in use.  

n etherin,out,err

number of packets sent and received per second, and total number of errors, displayed as separate graphs.  

p tlbpurge

number of translation lookaside buffer flushes per second.  

r reclaim

total pages of reclaimable memory. The graph displays the fraction of the machine’s total memory in use.  

s syscall

number of system calls per second.  

t tlbmiss

number of translation lookaside buffer misses per second.  

w swap

number of valid pages on the swap device. The swap is displayed as a fraction of the number of swap pages configured by the machine.  

8 802.11b

display the signal strength detected by the 802.11b wireless ether card; the value is usually below 50% unless the receiver is in the same room as the transmitter, so a midrange value represents a strong signal.  

z temp

current temperature reported by the cpu.  

The graphs are plotted with time on the horizontal axis. The vertical axes range from 0 to 1000*sleepsecs, multiplied by the number of processors on the machine when appropriate. The only exceptions are memory, and swap space, which display fractions of the total available, system load, which displays a number between 0 and 1000, idle and intr, which display percentages and the Ethernet error count, which goes from 0 to 10.. If the value of the parameter is too large for the visible range, its value is shown in decimal in the upper left corner of the graph.

Upper-case options control details of the display. All graphs are affected; there is no mechanism to affect only one graph.

-T sleepsecs

Set the number of seconds between samples to sleepsecs (default one second). Sleepsecs may be a floating-point number.  

-S scale

Sets a scale factor for the displays. A value of 2, for example, means that the highest value plotted will be twice as large as the default.  


Plot all graphs with logarithmic y axes. The graph is plotted so the maximum value that would be displayed on a linear graph is 2/3 of the way up the y axis and the total range of the graph is a factor of 1000; thus the y origin is 1/100 of the default maximum value and the top of the graph is 10 times the default maximum.  


If the display is large enough to show them, place value markers along the y axes of the graphs. Since one set of markers serves for all machines across the display, the values in the markers disregard scaling factors due to multiple processors on the machines. On a graph for a multiprocessor, the displayed values will be larger than the markers indicate. The markers appear along the right, and the markers show values appropriate to the rightmost machine; this only matters for graphs such as memory that have machine-specific maxima.  






Some machines do not have TLB hardware.