|
> > The problem with this policy is that it generates quite a bit of network
> > traffic between load balancer and the servers and therefore doesn't
> > perform well on highly loaded large systems.
>
> With the simulaton cluster I was trying feedbackd on (4 servers, 1
director),
> a health report once every 2 seconds was sufficient to prevent the 'herd
> effect'.
Hello,
Your strategy is then notably different from the AS that is discussed on
those papers. Their view is that the known load levels for the individual
servers must be very current for it to be useful in location policy for
individual jobs. If you consider a system with maybe hundreds or thousands
of service requests every second, then load level readings every two seconds
isn't what those papers mean with current. Please note that I quoted their
results on this, not mine.
Regards,
Marko
|
