All rights reserved. The Berkeley software License Agreement
specifies the terms and conditions for redistribution.
@(#)systat.1 6.3 (Berkeley) 05/02/85
While systat is running the screen is usually divided into two windows (an exception is the vmstat display which uses the entire screen). The upper window depicts the current system load average. The information displayed in the lower window may vary, depending on user commands. The last line on the screen is reserved for user input and error messages.
By default systat displays the processes getting the largest percentage of the processor in the lower window. Other displays show swap space usage, disk i/o statistics (a la iostat (1)), virtual memory statistics (a la vmstat (1)), network ``mbuf'' utilization, and network connections (a la netstat (1)).
Input is interpreted at two different levels. A ``global'' command interpreter processes all keyboard input. If this command interpreter fails to recognize a command, the input line is passed to a per-display command interpreter. This allows each display to have certain display-specific commands.
Certain characters cause immediate action by systat . These are
The following commands are interpreted by the ``global'' command interpreter.
Print the names of the available displays on the command line.
Print the load average over the past 1, 5, and 15 minutes on the command line.
Stop refreshing the screen.
Start (continue) refreshing the screen. If a second, numeric, argument is provided it is interpreted as a refresh interval (in seconds). Supplying only a number will set the refresh interval to this value.
Exit systat . (This may be abbreviated to q ).
The available displays are:
Display, in the lower window, those processes resident in main memory and getting the largest portion of the processor (the default display). When less than 100% of the processor is scheduled to user processes, the remaining time is accounted to the ``idle'' process.
Display, in the lower window, statistics about processor use and disk throughput. Statistics on processor use appear as bar graphs of the amount of time executing in user mode (``user''), in user mode running low priority processes (``nice''), in system mode (``system''), and idle (``idle''). Statistics on disk throughput show, for each drive, kilobytes of data transferred, number disk transaction performed, and average seek time (in milliseconds). This information may be displayed as bar graphs or as rows of numbers which scroll downward. Bar graphs are shown by default; commands specific to this display are discussed below.
Display, in the lower window, swap space in use on each swap device configured. Two sets of bar graphs are shown. The upper graph displays swap space allocated to pure text segments (code), the lower graph displays space allocated to stack and data segments. Allocated space is sorted by its size into buckets of size dmmin, dmmin*2, dmmin*4, up to dmmax (to reflect allocation policies imposed by the system). The disk segment size, in sectors, is displayed along the left hand side of the text, and data and stack graphs. Space allocated to the user structure and page tables is not currently accounted for.
Display, in the lower window, the number of mbufs allocated for particular uses, i.e. data, socket structures, etc.
Take over the entire display and show a (rather crowded) compendium of statistics related to virtual memory usage, process scheduling, device interrupts, system name translation cacheing, disk i/o, etc.
The upper left quadrant of the screen shows the number of users logged in and the load average over the last one, five, and fifteen minute interval. Below this line are statistics on memory utilization. The first row of the table reports memory usage only among active process, that is process that have run in the previous twenty seconds. The second row reports on memory usage of all processes. The first column reports on the number of physical pages claimed by processes. The second column reports the number of physical pages that are devoted to read only text pages. The third and fourth columns report the same two figures for virtual pages, that is the number of pages that would be needed if all processes had all of their pages. Finally the last column shows the number of physical pages on the free list. Below the memory display is the disk usage display. It reports the number of seeks, transfers, and number of kilobyte blocks transferred per second averaged over the refresh period of the display (by default, five seconds). For some disks it also reports the average millisecond per seek. Note that the system only keeps statistics on at most four disks. Below the disk display is a list of the average number of processes (over the last refresh interval) that are runable (`r'), in page wait (`p'), in disk wait other than paging (`d'), sleeping (`s'), and swapped out but desiring to run (`w'). Below the queue length listing is a numerical listing and a bar graph showing the amount of system (shown as `='), user (shown as `>'), nice (shown as `-'), and idle time (shown as ` '). At the bottom left are statistics on name translations. It lists the number names translated in the previous interval, the number and percentage of the translations that were handled by the system wide name translation cache, and the number and percentage of the translations that were handled by the per process name translation cache. Under the date in the upper right hand quadrant are statistics on paging and swapping activity. The first two columns report the average number of pages brought in and out per second over the last refresh interval due to page faults and the paging daemon. The third and fourth columns report the average number of pages brought in and out per second over the last refresh interval due to swap requests initiated by the scheduler. The first row of the display shows the average number of disk transfers per second over the last refresh interval; the second row of the display shows the average number of pages transferred per second over the last refresh interval. Below the paging statistics is a line listing the average number of total reclaims ('Rec'), intransit blocking page faults (`It'), swap text pages found in free list (`F/S'), file system text pages found in free list (`F/F'), reclaims from free list (`RFL'), pages freed by the clock daemon (`Fre'), and sequential process pages freed (`SFr') per second over the refresh interval. Below this line are statistics on the average number of zero filled pages (`zf') and demand filled text pages (`xf') per second over the refresh period. The first row indicates the number of requests that were resolved, the second row shows the number that were set up, and the last row shows the percentage of setup requests were actually used. Note that this percentage is usually less than 100%, however it may exceed 100% if a large number of requests are actually used long after they were set up during a period when no new pages are being set up. Thus this figure is most interesting when observed over a long time period, such as from boot time (see below on getting such a display). Below the page fill statistics is a column that lists the average number of context switches (`Csw'), traps (`Trp'), system calls (`Sys'), interrupts (`Int'), characters output to DZ ports using pseudo-DMA (`Pdm'), page faults (`Flt'), pages scanned by the page daemon (`Scn'), and revolutions of the page daemon's hand (`Rev') per second over the refresh interval. Running down the right hand side of the display is a breakdown of the interrupts being handled by the system. At the top of the list is the total interrupts per second over the time interval. The rest of the column breaks down the total on a device by device basis. Only devices that have interrupted at least once since boot time are shown.Display, in the lower window, network connections. By default, network servers awaiting requests are not displayed. Each address is displayed in the format ``host.port'', with each shown symbolically, when possible. It is possible to have addresses displayed numerically, limit the display to a set of ports, hosts, and/or protocols; see the list of commands below.
Commands to switch between displays may be abbreviated to the minimum unambiguous prefix; for example, ``io'' for ``iostat''. Certain information may be discarded when the screen size is insufficient for display. For example, on a machine with 10 drives the iostat bar graph displays only 3 drives on a 24 line terminal. When a bar graph would overflow the allotted screen space it is truncated and the actual value is printed ``over top'' of the bar.
The following commands are specific to the iostat display; the minimum unambiguous prefix may be supplied.
The following commands are specific to the vmstat display; the minimum unambiguous prefix may be supplied.
The following commands are common to each display which shows information about disk drives. These commands are used to select a set of drives to report on, should your system have more drives configured than can normally be displayed on the screen.
The following command is specific to the netstat display; the minimum unambiguous prefix may be supplied.
The remaining commands are common to displays which report network connections (currently only the netstat display). These commands may be used to select a specific set of connections for systat to report on.
/vmunix for the namelist /dev/kmem for information in main memory /dev/drum for information about swapped out processes /etc/hosts for host names /etc/networks for network names /etc/services for port namesAUTHORThe unknown hacker. The pigs display is derived from a program of the same name written by Bill Reeves.BUGSTakes 2-10 percent of the cpu. Certain displays presume a 24 line by 80 character terminal. The swap space display should account for space allocated to the user structure and page tables. The vmstat display looks out of place because it is (it was added in as a separate display rather than create a new program).The whole thing is pretty hoakey, it was included in the distribution under serious duress.