LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

RE: A new load balancing algorithm for discussion and consideration

To: "'Wensong Zhang'" <wensong@xxxxxxxxxxxx>, <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: RE: A new load balancing algorithm for discussion and consideration
From: "Marko Buuri" <marko@xxxxxxxxxx>
Date: Wed, 7 May 2003 09:02:28 +0300
>In the WLC algorithm, a new incoming job will not be counted in the 
>comparison of jobs/weight. So, in your example, the former 
>server has the 
>value (0/1) and the latter one has the value (1/3), then the 
>former server 
>is selected for the new job.

Hi,

Actually according to the URL below, alive connections Ci for each
server i is always 1 or more in WLC scheduling. Therefore I assume the
scheduling works like classic SED policy which is defined for server i
as SED(i) = (ni + 1) / ui, where ni is number of current jobs in server
i, and ui is service rate for server i. Note that because of the
incrementation by 1, the dividend is never less than 1. Just like it
appears to be for WLC scheduling as well.


URL: http://www.linuxvirtualserver.org/docs/scheduling.html


>I don't know whether your "Never Queue" scheme will consider 
>the server 
>weights or not. If not, it is probably similar to the Least-Connection 
>Scheduling.

Never Queue works just like SED (above) exept that any idle server is
selected first if available.


Marko

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