General discussion


DOWNLOA Achieve better Linux server performance

By Mark W. Kaelin Editor ·

What system performance monitoring tools do you use for measuring workload on your Linux servers? Is there a particular tool that you would like to see examined and explained?

This conversation is currently closed to new comments.

Thread display: Collapse - | Expand +

All Comments

Collapse -


by Jaqui In reply to DOWNLOAD: Achieve better ...

a statement made by a friend of mine that was teaching networking in a Utah University:

running red hat 8
serving an average of 18 gigabytes of data a day, peaking at 30 gigabytes

average cpu usage was 5%
peak was 7%

this was for a study he was conducting on data transfer usage with the 3d graphics community.
he had 20 sites hosted.

the peaks came when a client uploaded a new free download item.

he was planning on writing it up, but I don't know if it was ever done. I'll ask him.

Collapse -


by Jaqui In reply to hmmm

he hasn't written it up yet. he's looking for the materials.
( got a divorce, so moved. a lot of stuff just got packed away in a rush )

Collapse -

on unix - there's a 'top' command

by denwasson In reply to DOWNLOAD: Achieve better ...

I use the 'top' command at work on an HP-UX system, but I guess I haven't tried it on my linux box at home, so I don't know if it is available. It shows the processor load, and the top processes running. You can page up/down too.

Collapse -

yes . . .

by apotheon In reply to on unix - there's a 'top' ...

The top command is indeed available on Linux.

Collapse -

top and ps -A and uptime

by jmgarvin In reply to DOWNLOAD: Achieve better ...

I typically use top and uptime to keep an eye on usage.

I just take a look and see where the CPU and mem usage is out (via a script). If it goes to high, then I come in and make sure something funny isn't going on.

Collapse -

Good overview of available standard tools

by stress junkie In reply to top and ps -A and uptime

Edit: This was supposed to be a primary post; not a reply to jmgarvin's post.

The only tool in the article that I had not been using regularly is sar. Otherwise I've found that all of them have their place. iostat is great if you have a lot of disk activity and want to see the details. vmstat is great for lots of job control information. I'm surprised that netstat wasn't mentioned. It can be pretty handy once you find the options that you want to pass to it. And, as the article says, you can see a lot just by typing the contents of the pseudo-files in the /proc directory such as /proc/<ID>/cmdline. Big security hole though.

I always read these articles hoping that I will find something more akin to some of the tools that I have/had on DEC VMS. The VMS monitor utility alone was worth all of these Unix tools and more. Then the VMS command "show" was extremely overloaded with all kinds of things that you could watch.

The only thing wrong with VMS was that the developers didn't keep it current with trends in other products. The Motif GUI stunk and cost a lot of money. There were very few office type applications. The TCP/IP stack was late in coming and was implemented poorly. But as far as the kernel and security were concerned it was great. The system monitoring tools that came standard with the OS were great and you could tune the **** out of the system potentially resulting in enormous performance benefits.

I want my VMS back.

Related Discussions

Related Forums