[NBLUG/talk] Stress Testing servers

ME dugan at passwall.com
Thu Nov 13 10:14:00 PST 2003


You could quickly place a load of messages to be delivered by taking two
servers that are similar in configuration and setup test account that have
.forward to process each message and forward the e-mail back to the first
server sending account which also has a .forward to forward to 2 more
accounts.

1 message becomes 2 becomes 4 becomes 8 become 16 becomes 32 become 64...

In this way, you could quickly push the servers, in an increasing fashion
and maybe start a logger to show system performance per second and load as
this progresses. You could then take the two logs and compare when the
system felt unusable and examine the number of messages flowing through
per minute at that point.

This would give you an effective max capacity of messages per minute.

Then you could do this with messages of various sizes and see how that
alters peformance and capacity.

I think mail messages include notes on sending servers when forwarding
though. I think there is an inspection for relaying hosts and forwarding
hosts to prevent looping.If this is the case, you may need to munge
headers before forwarding messages.

Another method is to pubslish many, many e-mail addresses all over usenet
and sell them to spammers and let them provide the incoming data to stress
test your server. ;-)


Steve said:
> I just installed a new mail server for the company I work for.  It's not
> live yet =) But I want to stress test it big time for a couple days before
> it goes live.
>
> Anyone know of a way to stress test a server?  In this case I want to test
> the mail server (Sendmail) and make sure the box wont fall over under
> load.
>
> Any ideas are greatly appreciated.
>
> -Steve
>
>
> --
> ----
>       "Knowing others is wisdom, knowing your self is Enlightenment."
>                                                    -- Lao-Tzu
> _______________________________________________
> talk mailing list
> talk at nblug.org
> http://nblug.org/mailman/listinfo/talk
>
>




More information about the talk mailing list