At 11:45 AM 10/16/02 -0700, John Bellardo wrote:

>>> As best I can determine all our major platforms support the ability to
>>> send sockets between processes, so this "load balancer" has a constant
>>> (close to single context switch) overhead.
>>>
>>
>
>I don't know Win32, but check "man 7 unix" on a linux system. There is
>a note in the man page that this API is not necessarily portable, but
>I've seen similar offerings on other systems.
>

Good grief! How do they expect somebody to find that? What
does "man 7 VMS" and "man 7 Windows XP" do?

A load balancer would have a clear advantage over pure classic
as process creation, initialization, database open, and metadata
acquisition would only be necessary once per worker process
rather than once per attachment. Huge win over classic. Couldn't
compete with thread-friendly superserver under load, however.
No-IPC is always faster that IPC (duh!). But it would almost
certainly whomp current superserver on SMP.

Thanks for the pointer (old wolf learns new trick).

Jim Starkey