|
Edward Chee wrote:
>
> I see. Pls excuse my ignorance - I'm new to all this :-)
no problem. I have to think about this too.
> Hmm. I wonder how this TCP port issue directly limits the Zeus software load
> balancer's incoming connection rate. In section 1.4 of
> http://support.zeus.com/doc/zlb/sizing_guide.pdf, they describe this
> problem.
We don't hear much about commercial software here, I assume because either
people don't want to spend the money or they want the source code. Someone
here may know how it works. I do know the time I thought of entering an LVS
in the Polygraph torture tests, that an LVS would not connect into the test
setup. The Polygraph test bench required that all (real)servers have their own
IP,
and be unmodified (no VIP on lo say). The Polygraph people supplied the clients
and servers (all *BSD), you just brought in your load balancer. You could
fit an LVS-NAT director into this test setup, but
back in the 2.2 days, LVS-NAT was slow. The polygraph people did talk about
something called "direct server return" in which the packets returned directly
from the server to the client, but it wasn't like LVS-DR, since an LVS-DR
setup wouldn't work. I couldn't figure out how it worked and got the idea
that unless I knew what it was, that I really didn't belong there.
So if you figure out how the Zeus load balancer works, I'd be glad to hear
about it.
> To rephrase: A load balancer that operates on Layer 7 and maintains
> persistent connections to the real servers has to act as a TCP client and is
> subject to the port limitations. Is my thinking correct?
I guess so, but do you know that this is how it works?
Joe
--
Joseph Mack PhD, High Performance Computing & Scientific Visualization
SAIC, Supporting the EPA Research Triangle Park, NC 919-541-0007
Federal Contact - John B. Smith 919-541-1087 - smith.johnb@xxxxxxx
|
