LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: Off topic: Realserver performance.

To: Joseph Mack <mack@xxxxxxxxxxx>
Subject: Re: Off topic: Realserver performance.
Cc: lvs-users@xxxxxxxxxxxxxxxxxxxxxx
From: Benjamin Lee <benjaminlee@xxxxxxxxxxxxxx>
Date: Wed, 20 Sep 2000 01:19:14 +1100
On Tue, Sep 19, 2000 at 09:50:30AM -0400, Joseph Mack wrote:
> with LVS you don't have that choice. Each real-server is loaded
> progressivley attempting to keep the load the same on all machines.
> You can't cut off one machine at a certain load.


Yup, of course in my simulation I have a weighted polling queue and the servers 
can remove themselves from the queue by using an invalid weight. (And of 
course, it's only a simulation!)

As regards to LVS, I don't think it would be too tricky to implement a dynamic 
round-robin like scheduler in which the candidate for the round robin can 
invalidate themselves if they believe they are too heavily loaded.

The crux of what I am doing of course, which has already been brought up, is 
trying to gauge 'load'. 


-- 
B.      http://b.makelinux.org/      "Of course there is a reason why!"
__________________________________________________________________________
If you live long enough, you'll see that every victory turns into a defeat.
                -- Simone de Beauvoir


<Prev in Thread] Current Thread [Next in Thread>