NAME
osview - monitor operating system activity data
SYNOPSIS
osview [-in] [-nn] [-unamelist] [-s] [-c]
DESCRIPTION
osview monitors various portions of the activity of the operating system
and displays them using the full screen capabilities of the current
terminal.
A large number of activity counters are monitored, and the display may be
dynamically altered to hide or show only those counters in which the user
is interested. It is assumed that the osview user is somewhat familiar
with the internal workings of an AT&T V.4 based kernel.
osview lays out as much information as possible in the screen area
available. Each data item is grouped similarly to the grouping shown by
sar(1). A header line gives the group name, and members of the group are
indented below along with the one-second average value over the last
interval (or total value over the interval; see below). If a graphics
subsystem is not present on the machine being monitored, osview
suppresses all graphics related statistics in the display.
The -i parameter sets the delay between screen updates in seconds. By
default, a 5 second rate is used. The -n parameter is used to override
the default number of lines to use, which is usually the entire size of
the display area. This can be useful if the display is called up in a
long window, to keep the counters grouped together at the top of the
window. The -s option informs osview to not reduce relevant values to
the average over a second. One second averaging allows instant
performance estimates, but may show inaccuracies because of the short
interval involved. The -c option causes a running count to be displayed,
rather than an interval count. The counts can be reset to zero by
pressing the C key. Finally, the -u option allows the specification of a
different namelist file for those symbols which must be read from the
running kernel. By default, the normal namelist file /unix is used.
In general, those parameters dealing with data throughput rather than
events are presented as the number of bytes involved. For instance,
memory usage is reported in bytes, as well as buffer cache traffic.
Those parameters dealing with events to the system, such as page fault
activity, interrupts or system activity are reported as actual counts.
This allows an instant estimate of the activity and throughput of the
system.
A group can be suppressed along with all its members to allow hidden
groups to be brought into view if the screen area is too small. This is
done by moving the cursor over the header line of the group to suppress
and typing a suppression character. The cursor may be positioned in any
of the standard ways; keyboard arrow keys, the h-j-k-l keys, or the
backspace-return-tab keys. osview highlights the line the cursor is on
unless the cursor is on the top screen line (which is reserved for status
information). When positioned over a group name, typing the D character
or one of the delete keys on the keyboard will suppress the group. The
group name will remain, with an asterisk (*) prefix to indicate that the
group has been suppressed. The group may be expanded again by
positioning the cursor over the group name and typing the I character or
one of the insert keys on the keyboard. The home key moves the cursor to
the osview status line.
OVERVIEW
The information which osview displays and how to interpret it is given
below. See the documentation for sar(1) or gr_osview(1) for additional
information. Some headers, including Swap, and Interrupts are suppressed
by default. See above description of how to get them to display. Some
headers, including PathName Cache, EfsAct, and Getblk contain information
that is subject to change and is of use primarily by IRIX development
groups.
Load Average
These counters give a Tenex-style load average over the last minute, 5
minutes or 15 minutes.
CPU Usage
These counters display the proportion of the available processor
cycles which were used by each of the following activities. If
multiple processors are present, than the CPU number will be added to
the header line.
user - user programs
sys - system on behalf of user
intr - interrupt handling
gfxc - graphics context switching
gfxf - waiting on graphics input FIFO
sxbrk - waiting for memory
idle - doing nothing
Wait Ratio
%IO - waiting on IO
%Swap - waiting on swap IO
%Physio - waiting on physical IO
Real Memory
Phys - physical memory size
Kernel - memory consumed by kernel text and data
Heap - part of Kernel used by heap
Stream - part of Heap used by streams
Zone - part of Kernel used by zone allocator
Ptbl - part of Kernel used by process page tables
Fs Ctl - memory holding filesystem meta-data
Fs Data - memory holding filesystem file data
Delwri - modified filesystem file data
Free - memory not in use
Userdata - in use holding valid user data
Freeswap - physical swap space available
Vswap - virtual swap space and swappable memory available
Pagealloc - physical pages allocated from free pool
Virtual Memory
vfault - page faults
pfault - protection faults
demand - demand zero and demand fill faults
cw - copy-on write faults
steal - page steals
onswap - page found on swap
oncache - page found in page cache
onfile - page read from file
freed - pages freed by paging daemon
unmodswap - clean swap page, dirty incore page
unmodfile - clean file page, dirty incore page
Iclean - number of icache cleans
Block Devices
lread - amount of logical buffer reads
bread - amount of physical buffer reads
%rcache - read hit ratio on buffer cache
lwrite - amount of logical buffer writes
bwrite - amount of physical buffer writes
wcancel - amount of delayed writes cancelled
%wcache - write hit ratio; negative for write-behind
phread - amount of raw physical reads
phwrite - amount of raw physical writes
TLB Actions
newpid - new process ID allocated
tfault - second level TLB misses
rfault - reference faults (during paging)
flush - flush of entire TLB
sync - cross-processor TLB synchronizations
Graphics
griioctl - graphics ioctl's
gintr - graphics interrupts
swapbuf - swapbuffer completes
switch - context switches
fifowait - wait on FIFO
fifonwait - wait on FIFO, below low-water mark on check
System Activity
syscall - system calls
read - read system calls
write - write system calls
fork - fork system calls
exec - exec system calls
readch - characters read via read()
writech - characters written via write()
iget - efs inode searches
TCP
sndtotal - packets sent
rcvtotal - packets received
sndbyte - bytes sent
rcvbyte - bytes received
Net IF
These counters display the activity on a particular network interface.
If multiple interfaces are present, than a separate set of counters is
displayed for each interface. The interface name is displayed as part
of the header.
Ipackets - packets received
Opackets - packets transmitted
Ierrors - packets received in error
Oerrors - errors transmitting a packet
collisions- collisions detected
Scheduler
runq - number of processes on run queue
swapq - number of processes on swap queue
switch - context switches
Interrupts
all - total interrupts handled
vme - VMEBus interrupts
Swap
swapin - page swapins
swapout - page swapouts
bswapin - bytes swapped in
bswapout - bytes swapped out
SEE ALSO
gr_osview(1), top(1), sar(1).
BUGS
osview cannot atomically get all the data it needs. On a very busy
system, some percentages could sum to greater than 100, since there could
be a gap between the time osview reads the current time and when it reads
the data counters.