Re: IceBreak as an alternative to Apache?

[Thorbjoern Ravn Andersen <ravn@xxxxxxxxxx>]

Den 01/01/11 15.45, Aaron Bartell skrev:
> I am in over my head with talking about the network TCP/IP layer, but
> essentially, yes, that is what I was alluding to.  I am more repeating what
> IBM told me than understanding all of the implications of what they said.
No problem.  I am just interested in learning what limitations you may 
have encountered and under what circumstances to prepare for them.

If it is the buffer in the network layer that got full, I agree with the 
IBM developer that you should not increase the buffer since this does 
not solve the actual problem, namely that your programs do not accept 
new incoming connections fast enough to keep up.

> I dont' have access to a 7.1 machine on my LAN, though I could spin up an
> IBM Virtual Loaner Instance.
Might be quite interesting, but perhaps not if the problem is in your 
code. Perhaps you want to create a "just return a tiny response 
immediately" to the accept code, to see if the i can keep up for longer 
with your load inducer on your local network if you do not do any 
processing.