Overview of Existing Systems and Unix Specifics

There are so few systems that provide dynamic load balancing that little has been done to enhance the basic monitoring present on each central system by providing remote access to each system's load monitoring through Remote Procedure Call Service to that system. Unix/solaris tools for local performance monitoring include:

ps - process status listing (related public domain programs such as top and mon provide continual versions of same info - usually by reading /dev/kmem)
vmstat - Virtual Memory statistics - see for example #tbvmstat#2492>
#table2493#
Table: Virtual Memory Usage example output
iostat - Input Output statistics - typically, as in #tbiostat#2496>.
#table2497#
Table: Input/Output example
netstat Can shows configuration, status and simple statistics
lpstat
uptime (the UCB version) Shows the first line of ;SPM_quot;w;SPM_quot; output, including the average number of jobs in the ;SPM_quot;ready;SPM_quot; queue over the last sample periods of 1, 5 and 15 minutes - This is in fact one of the MOST useful characterizations of a system that is available.

Network versions of some of these are available through Sun RPC servers:

ruptime - retrieve the process statistics from remote systems.
rup - retrieve system load characteristics.
rusers - return information about users' processes.
rstat/statd - return statistics from remote load server.

For process and system characterisation, there are tools for user and system profiling:

user space profiling prof/gprof monitor via the trace system call etc.
kernel profiling, through system administrators interfaces such as sad.