What's the easiest way to display system cpu, memory, # drives/size ??

UNIX for Advanced & Expert Users
1

I am looking for the easiest and most generic way
to determine:

System model/class
Number of cpu's
Clock speed of cpu's (ie 550 MHz)
Total Physical Memory (not virtual)
Number of Drives/Drive Size

Thanks in advance,

2

For what OS are you looking to get this complete information? For linux if you installed KDE or GNU they include a user friendly Xsystem panel for you to check those information called "Process Managment".

  • For solaris mysys> prtconf | grep Mem for checking on the Memory size, top command, primarily used to identify a system's most resource-consuming processes, also provides a synopsis of memory and memory usage.

  • If you still unable to figured out what you are looking for, then check your MAN page depending on what UNIX os you are using the commands might vary..

3

in Solaris, you may try

/usr/platform/YourMachineModel/sbin/prtdiag

i.e. /usr/platform/sun4u/sbin/prtdiag
or /usr/platform/SUNW,Ultra-1/sbin/prtdiag
hope this help

:wink:

4

You can also pull a lot of that information out of the `dmesg` output.

That is normally the first place that I look...

5

There you probably have the only real way to do it.
I have never seen that SunOS stuff on any other system, and if I should vote between that and GNU /proc, I'd vote for the latter.

But in dmesg you find what the system found, and dmesg is on all systems I know about.

Could you tell us what you want this information for?

These are the commands I found that deal with some of this:

uname - Will tell you about the system name
top - You may parse some of this stuff and
draw inferences from it
df - This tells you about drives and free
space (parse for NFS mounts!)

To determine system speed, you have to make a small program, I guess.
Time a loop or something, maybe invalidate the cache inside the loop, determine what you want as criterion, and divide that by the time it took to run.

double d, dd;
long i;
long t = \(long\)time\(\);
d = \(double\) t;

for (i=0; i < LOOPCNTR; i++)
invalidate_cache();
t = (long) time();
dd = (double) t;

time\_it_took = dd - d;

system_speed = criterion / time\_it_took;

You could fork one for each 'suspected' process and time it again. With a few if's - like each new process is given a new processor, you may infer the number of processor by multiplying time_it_took by the number of forks and compare.

This is a BFAMI(*) approach that might work.

Physical memry can be determined with a

long *p = 0;
while\(p\+\+\);

loop that evenually will seqfault.
Trap that, and check the count.
Another BFAMI approach.

A structured approach is write to the POSIX guys and tell them: "Hey, I think we need a portable way of determining the system configuration, like CPUs, memory, disks, etc."

Maybe they'll listen.

Write an RFC.

Pray.

Atle

(*)(BFAMI = brute-force-and-massive-ignorance)